J O I N T C E N T E R

AEI-BROOKINGS JOINT CENTER FOR REGULATORY STUDIES

Working Paper 03-4

June 2003

Thomas J. Kniesner is the Krisher Professor of Economics and Senior Research Associate, Center for

Policy Research at the Maxwell School of Citizenship and Public Affairs. John D. Leeth is Professor

of Economics at Bentley College. The authors are grateful to the Mine Safety and Health

Administration for providing the data necessary to complete this study. They are especially grateful to

Fred Menke of MSHA for helping with their request for data and for answering their questions about

MSHA procedures. Finally, the authors want to thank Dan Black, Gary Engelhardt, Bill Horrace, Jeff

Kubik, Jeff Racine, John Moran, the members of Syracuse University’s Economics Department

Seminar and Syracuse University’s Public Finance Breakfast, and the members of the Health and

Safety Seminar of the McDonough School of Business at Georgetown University for their helpful

comments. All communications can be sent to Thomas J. Kniesner; Center for Policy Research, 426

Eggers Hall; Syracuse University; Syracuse, NY 13244–1020 USA; (315) 443–4589; Fax: (315) 443–

In response to growing concerns about understanding the impact of regulation on

consumers, business, and government, the American Enterprise Institute and the

Brookings Institution have established the AEI-Brookings Joint Center for Regulatory

Studies. The primary purpose of the center is to hold lawmakers and regulators more

accountable by providing thoughtful, objective analysis of existing regulatory programs

and new regulatory proposals. The Joint Center builds on AEI's and Brookings's

impressive body of work over the past three decades that has evaluated the economic

impact of regulation and offered constructive suggestions for implementing reforms to

enhance productivity and consumer welfare. The views in Joint Center publications are

those of the authors and do not necessarily reflect the views of the staff, council of

academic advisers, or fellows.

ROBERT W. HAHN ROBERT E. LITAN

Executive Director Director

KENNETH J. ARROW

Stanford University

MAUREEN L. CROPPER

University of Marylandand World Bank

PHILIP K. HOWARD

Covington & Burling

PAUL L. JOSKOW

Massachusetts Institute of Technology

DONALD KENNEDY

Stanford University

ROGER G. NOLL

Stanford University

GILBERT S. OMENN

University of Michigan

PETER PASSELL

Milken Institute

RICHARD SCHMALENSEE

Massachusetts Institute

of Technology

ROBERT N. STAVINS

Harvard University

CASS R. SUNSTEIN

University of Chicago

W. KIP VISCUSI

Harvard University

All AEI-Brookings Joint Center publications can be found at www.aei.brookings.org

© 2003 by the authors. All rights reserved.

Studies of industrial safety regulations, OSHA in particular, often find little effect on

worker safety. Critics of the regulatory approach argue that safety standards have little to

do with industrial injuries, and defenders of the regulatory approach cite infrequent

inspections and low penalties for violating safety standards. We use recently assembled

data from the Mine Safety and Health Administration (MSHA) concerning underground

coal mine production, safety regulatory activities, and workplace injuries to shed new

light on the regulatory approach to workplace safety. Because all underground coal mines

are inspected at least once per quarter, MSHA regulations will not be ineffective because

of infrequent inspections. We estimate over 200 different specifications of dynamic mine

safety production functions, including ones using deliberately upward biased estimators,

and cherry pick the most favorable mine safety effect estimates. Although most estimates

are of insignificant MSHA effects, we select the single regression specification producing

the most favorable MSHA impact from the agency viewpoint, which we then use in a

policy evaluation. We address the question of whether it would be cost-effective to move

some of MSHA’s enforcement budget into alternative programs that could also improve

the health of the typical miner. Even using cherry picked results most favorable to the

agency, MSHA is not cost effective at its current levels. Even though MSHA is a small

program when judged against others like OSHA and EPA, MSHA's targeted public

health objective could be much better served (almost 700,000 life years gained on

balance for typical miners) if a quarter of MSHA's enforcement budget were reallocated

to other programs such as more heart disease screening or defibrillators at worksites.

1

There is much evidence that OSHA inspections have not been effective in reducing

injuries (Kniesner and Leeth 1995, Chapters 1 and 2). One explanation is that there are

too few OSHA safety inspections to make a difference to firms. OSHA inspections also

might not matter much to worker safety because of low fines or irrelevant safety

regulations. In contrast to the relatively infrequent OSHA inspections in construction or

manufacturing, mines regulated by the Mine Safety and Health Administration (MSHA)

are inspected quarterly. Also in contrast to OSHA inspections, mine safety inspectors can

effect a work stoppage until a safety violation is corrected. We use recently assembled

data on underground coal mine production, injuries, and safety inspection and other

regulatory activity to examine econometrically the effectiveness of the regulatory

approach to workplace safety where the law is potent and inspections frequent. We find

that even if we cherry-pick results to maximize the estimated effectiveness of MSHA

there is an excess of inspections in mining, which has a notable cost of foregone

opportunities to improve the typical miner’s health through other existing means.

By way of background, the Federal Coal Mine Health and Safety Act of 1969, as

it is formally called, was the most comprehensive and stringent Federal legislation

covering the mining industry. It included surface as well as underground coal mines,

required two annual inspections of every surface coal mine and four at every

underground coal mine, and greatly increased federal enforcement powers in coal mines.

The Coal Act required monetary penalties for all violations and established criminal

penalties for knowing and willful violations. The safety standards for all coal mines were

strengthened and health standards adopted. The Coal Act included specific procedures for

developing improved mandatory health and safety standards and established

compensation for miners who were totally and permanently disabled by the by the

progressive respiratory disease know as black lung. The annual enforcement budget of

2

MSHA is about 60 percent the size of OSHA’s, $110 million. (For more details see

http://www.msha.gov).

A focal point of our research is to estimate an econometrically sophisticated

regression model of the connection between mine inspections and mine safety outcomes.

We adopt the general dynamic panel model of Arellano and Bond (1991), which

incorporates sluggish adjustment between desired safety outcomes along with

endogeneity of both production and safety policy at the mine level. We purposely

examine a large number (200+) of econometric specifications, including ones deliberately

biased upwards, so as to find maximal MSHA effects.

The dynamic quarterly unbalanced panel model we estimate is in sharp contrast to

the empirical specifications in the existing literature. Previous work examines aggregate

trends in injuries without covariates directly related to MSHA activities and attributes

success to MSHA if a downward trend in coal mine injuries continued or accelerated

after MSHA (Lewis-Beck and Alford 1980, Weeks 1995) or infers a positive effect of

MSHA on mine safety if injuries are lower in the post-MSHA period without any

consideration of the pre-MSHA pattern of injuries (Neumann and Nelson 1982, Fuess

general background trend in mine injuries but also because do not simply attribute

unexplained changes in injuries to mine safety regulation as we have direct measures of

safety regulation enforcement.

To summarize our results, we focus on estimates of the exogenous general

deterrence effects of MSHA, which capture regulatory activities for the mine’s

enforcement district. In only 1/200 specifications are the estimated MSHA effects on

injuries negative and large relative to their standard errors. Purposely ignoring the issue

that statistical significance is suspect when the data have been mined ex ante (Lovell

1983), we then use the single set of parameter estimates most favorable to MSHA in a

policy evaluation of the agency’s current regulatory activities. Even if a modest amount

of MSHA’s relatively small enforcement budget, say 25 percent, were reallocated to

other public health programs targeted to the demographic groups that are typically

3

miners, there would be a substantial gain in health status of the target population (about

700,000 additional life years).

We begin by describing the information needed to examine the cost-effectiveness

of mine safety policy. We first answer the focal question of our research. What do we

need to do econometrically with our newly constructed data set on underground coal

mines to estimate the response parameters required for evaluating the workplace safety

effects of MSHA?

Effective safety inspections can reduce deaths and injuries simultaneously. If

totinj is the total number of injuries (fatal and non-fatal) then the algebraic expression for

the economic benefit (B) of reducing one workplace (mining) injury is the economic

In (2.1) VL is the revealed value of life, VI is the revealed value of avoiding injury, and

coal mining. The value of injury reduction is a weighted combination of the values placed

on avoiding fatal and non-fatal injuries.

Let wonum be the number of inspections per mine per quarter with at least one socalled

withdrawal order (the mine must remove workers from the mine) because of a

serious safety or health violation. Next, let pennum be the number of inspections per mine

per quarter with at least one monetary fine for a serious safety or health violation. If

MSHA inspections neither ignore dangerous conditions nor concoct ones that do not exist

then the only way for MSHA to improve safety in underground coal mines is to increase

the number of inspections. The additional inspections would then result in more

withdrawal orders or more monetary penalties per mine.

If I is the number of inspections per mine then the maximum safety impact of

additional inspections through additional withdrawal orders and monetary penalties is

4

( )

¶ ¶ ¶ ¶

=- -

¶ ¶ ¶ ¶

. (2.2)

In the case where the proportions of withdrawal order inspections and monetary penalty

¶ ¶

=- -

¶ ¶

. (2.3)

Because diminishing returns to inspecting mines may hold in reality, (2.3) produces a

lower bound to the number of inspections it would take to eliminate one injury, which in

turn means that calculations based on (2.3) may overstate the cost-effectiveness of

MSHA, giving the benefit of the doubt to the agency.

The most important aspect of our research will be the safety outcomes regressions

solve for dI to determine the number of additional MSHA inspections needed to eliminate

one workplace injury, which is

( ) ( )

1

¶ ¶

=

+

. (2.4)

To address the issue of cost effectiveness, we can compare the cost of the

additional inspections evaluated in (2.1).

It is helpful to place into an economic context of the firm the two partial

derivatives in (2.4) that are the primary components of the cost of MSHA safety

enforcement activities. The small-scale economic model of the firm we present clarifies

how to estimate econometrically the effectiveness of MSHA in a way that improves on

the existing empirical literature concerning the cost-effectiveness of the regulatory

approach to enhancing mine safety.

Consider a mine in year t with an (endogenous) optimal stock of health and safety

5

In the typical situation safety capital is productive and regulation not counterproductive

In the econometric specification of the inverse safety production function that we

estimate we acknowledge that the impact of MSHA enforcement involves multiple

activities, each of which can have non- linear effects described by the reverse S-shape

depicted in Figure 1. In particular, we will allow the starting level of MSHA enforcement

to condition its marginal effectiveness, such that more enforcement may be ineffective

when starting from either very low or very high initial levels. The shape in Figure 1 also

implies that reducing MSHA somewhat from very high initial levels can be costeffective.

Before we describe the econometric model and the resulting parameter estimates

there are a few more conceptual details to flesh out in (3.1). First, safety capital wears

be conditioned by the characteristics of workers and the technology contributing to

represent the other econometrically includable conditioning factors, the inverse safety

production function is

The optimal amount of safety capital, q, at time t depends on its previous level

and the firm’s desired investment in safety capit al, both of which depend on previous

about the regression specifications to follow, the model is econometrically indistinguishable from the

behavioral regulation approach (Scholtz and Gray 1990) and the optimizing social regulator approach

(Auld et al. 2001).

6

(after allowing for threshold effects depicted in Figure 1) MSHA has its intended effects

on workplace activities. Because we do not have direct observations on investment in

effect on safety plus any indirect effect through the agency’s impact on health and safety

The theoretical discussion of the last section emphasized the need for control

covariates and dynamic adjustment to the ultimate equilibrium safety level, which leads

naturally to the Arellano-Bond dynamic panel model that is summarized in general

algebraic form as

1 2

1

=

2

inverse safety production function of (3.2) with the Arellano-Bond estimator of (4.1) the

dependent variable is a mine’s total injuries; the predetermined variables include

production levels and mine-specific MSHA enforcement activities, and exogenous

on later realizations of w, then w is a predetermined variable. The idea is that unforecastable errors today

might affect future changes in w.

7

variables include mine district MSHA enforcement activities plus mine district and time

dummies.

The Arellano-Bond estimator proceeds by first differencing (4.1), which removes

lagged levels of the dependent variable and predetermined variables and differences of

A practical problem with the Arellano-Bond estimator is that predetermined

variables greatly increase the size of the instrument matrix. A very large instrument

matrix makes GMM estimators perform poorly in small samples or makes the model

It also is important to note that there are two versions of the Arellano-Bond

estimator, a one-step estimator and a two-step estimator, which adds additional

complexity for the applied researcher. In the one-step estimator the Sargan test overrejects

the overidentifying restrictions when there is heteroskedasticity. However, the

standard errors of the two-step estimator are biased downward in small samples. So, the

researcher generally uses both the one-step and two-step Arellano-Bond estimators, but

for different purposes. The two-step results are better for model specification testing of

the over- identifying restrictions, and the one-step results are better for inferences on the

regression coefficients. Finally, it is important to note that the dynamic model in (4.1) is

not identified if the dependent variable is persistent (a pure random walk makes lagged

second-order autocorrelation in the first-differenced errors; the XTABOND routine produces so-called

robust (to heteroskedasticity) standard errors, incorporates the needed tests for autocorrelation as well as

the Sargan test of the overidentifying restrictions. The Sargan test is poorly sized, however, and difficult to

pass when the instrument set is relatively large as in our case (Hall and Horowitz 1996, Ziliak 1997); it is

therefore not surprising that none of the regressions we discuss pass the Sargan test.

which adds another column to the instrument matrix Z, and so on, which are in addition to the columns for

each x. More generally, if p is the number of lagged y’s in the model, i is the number of cross-section units,

variables are like lagged y’s in terms of adding columns to the instrument matrix. In our estimation we

work with 1–8 lags of y, 1–3 predetermined variables, 1–206 x’s, with the maximum T = 55 and i = 3450,

so that our models are often constrained by the maximum feasible width of Z in STATA.

8

levels of y weak instruments, and weak instruments lead to finite sample bias in panel

To generate our data for estimation we merge five separate data sets provided by

the Mine Safety and Health Administration covering inspections, violations, assessed

penalties, injuries, and production and employment. The five data sets provide unique

tracking numbers for each inspection, violation, and mine. Using the violations and

inspections tracking numbers we link the information on assessed penalties and violations

to information on inspections. We then combine enforcement information and quarterly

data on production and employment based on mine identification numbers and beginning

dates of the inspections. Likewise, we tie injury information to each mine and each

we use in estimation contains quarterly information on MSHA enforcement efforts and

mining injuries, employment, and production during 1983–1997.

Although MSHA enforcement efforts may have an immediate effect on the

frequency or severity of accidents, they are unlikely to change the immediate incidence of

health-related problems such as hearing loss or black- lung disease. Mine-related diseases

develop gradually so that it is unlikely we can adequately determine the effect of MSHA

enforcement efforts on miner health using information spanning the 15 years available.

Accordingly, we exclude from the original data set all inspections focusing on health

(such as inspections of a mine’s ventilation system or monitoring for dust, noise or

silica). To narrow our focus further to inspections likely to improve miner safety directly

we also exclude all MSHA actions not on mine property, activities related to education

and training, investigations for discrimination, and audits of accident, injury, illness, and

employment records. In every case where we exclude an inspection or MSHA activity we

likewise exclude the resulting citations, orders, and penalties.

we exclude outside contractors from our research.

9

Figures 2–4 depict the history of mine safety, including the quarterly data in our

estimation sample. What the annual data in Figure 2 show is that MSHA, as it post-dates

the 1969 Coal Act, may have had its intended effect of improving safety, at least where

fatal injuries are concerned. Figure 3 also reveals the possibility that non- fatal injuries too

may be affected by MSHA, most recently since the mid-1980s. Finally, Figure 4, which

plots the quarterly data in our estimation sample of 1983–1997, emphasizes the

seasonality of injuries as well as supports the possibility that MSHA has been effective in

reducing miner injuries since the middle 1980s.

Table 1 presents summary statistics on MSHA safety-related enforcement

been adjusted to reflect inflation to 2002. The first two panels provide information on

individual citations and orders to withdraw miners from the worksite, the third panel

reports MSHA penalties per inspection, and the last two panels provide total enforcement

efforts per quarter.

Panels A and B of Table 1 reveal that most, but not all, MSHA penalties were

imposed for serious violations of health and safety standards. About 60 percent of

citations and 57 percent of withdrawal orders were issued for violations that MSHA

inspectors viewed as significant and substantial, or likely to result in injury.

With an average of $184, initial fines on citations were fairly small. Fines on

withdrawal orders were considerably larger, averaging $2,079, although only 60 percent

penalties downward over time. Monetary penalties on withdrawal orders fell from their

initial amounts an average of 39 percent and monetary penalties on citations fell from

their initial amounts an average of about 14 percent.

The degree of operator negligence may at least partially explain the much larger

average fine on withdrawal orders than on citations. Approximately 88 percent of the

violations resulting in a withdrawal order were classified by MSHA inspectors as caused

by a high degree of operator negligence or reckless disregard of mine safety, whereas

citation.

1 0

only about two percent of the violations resulting in the issuance of a citation were

caused by a high degree of operator negligence or reckless disregard of miner safety.

Besides a monetary penalty, withdrawal orders also shut down production, which

imposes a potentially large cost if operations are disrupted for an extended period. At

least fifty percent of the withdrawal orders were terminated fairly quickly, in one day or

less, but for a sizable number the days from issuance to termination extended for weeks

and, in some cases, months and years. Because of the extremes the average number of

days from issuance to termination is large, 32.2 days. Appendix A provides additional

details of the distribution of lost days. As shown, 25 percent of all withdrawal orders

from 1983 to 1997 extended for 6 days or more and 5 percent extended for 112 days or

more. With such a potentially long shut down period, mines had strong incentives to

avoid conditions likely to result in withdrawal orders. Additionally, the harsh penalties

shown in Appendix A for failure to abate and imminent danger hazards likely motivated

mines to rectify previously discovered problems and avoid conditions liable to result in

death or severe injury. By way of contrast, the incentives to avoid citations resulting only

in monetary penalties were quite small. From 1983 to 1997, 95 percent of all citations

had initial fines less than $447 and 99 percent had initial fines less than $1,158. By law

mines must continue to pay miners for the remainder of the shift during which a

withdrawal order is issued and for up to 4 hours the next day if the withdrawal order is

still in effect, meaning that the monetary losses from a withdrawal order almost always

substantially exceed the losses from a simple citation resulting in a fine.

As can be seen in Panel C of Table 1, about 40 percent of MSHA safety

inspections led to a monetary penalty, and about five percent of inspections resulted in a

withdrawal order. About 31 percent of all inspections uncovered at least one serious

violation of MSHA health and safety standards, and about five percent of inspections

discovered at least one violation with a high degree of operator negligence. For

inspections where a monetary penalty was imposed the initial fines for all citations and

withdrawal orders issued during the inspection averaged $1,503. Over time, the monetary

penalties fell by about 14 percent on average.

Panel D of Table 1 presents MSHA enforcement efforts per quarter. The average

operating coal mine was inspected for safety-related problems 4–5 times per quarter but

1 1

had slightly less than two inspections per quarter resulting in monetary penalties and

0.211 inspections per quarter resulting in withdrawal orders. Per inspection with fines,

the average monetary penalty for all violations was $1,263. On average, MSHA reduced

monetary penalties by about 16 percent from initial le vels.

Finally, Panel E of Table 1 indicates quarterly enforcement efforts with all minor

violations excluded, which are violations that MSHA inspectors believe are unlikely to

result in injury. MSHA can improve miner safety to the extent that inspectors can

identify serious violations of safety standards, violations likely to result in injury. MSHA

discovered serious violations of safety standards in a large majority of mines each

quarter. Slightly more than 71 percent of all mines received at least one monetary penalty

for a serious violation of safety standards. Per quarter, the average operating coal mine

had about 1.4 inspections resulting in monetary penalties for serious violations, 0.11

inspections resulting in withdrawal orders, and 0.18 inspections with one or more high

degree of operator negligence violations. The average fine in a quarter for mines with

serious violations that receive monetary penalties was $1,442. Over time monetary

penalties fell on average by about 17 percent.

We now describe the large number of specifications of a dynamic mine safety

equation that we estimated. In contrast to Sala-I-Martin (1997) who examines several

million regressions to find the true model of country growth, we search among a large

number of regressions to find the single set of results the most favorable to MSHA. We

then use our shamelessly data mined results in best-case calculations of the cost

effectiveness of MSHA and it implications for improving the health and safety of the

population typically working as miners.

Our research also is similar to a meta-analysis on one data set because in the

process of estimating a large number of econometric specifications we will as a byproduct

see if a pattern emerges with regards to the effectiveness of MSHA in influencing

miner safety. The large number of regression specifications (200+) comes about because

we consider various (1) safety measures (total injuries, injury rate, fatal injuries, non-fatal

injuries, zero versus some injuries), (2) MSHA activities (specific abatement, general

1 2

deterrence, both), (3) instrument sets (small, medium, large), (4) time frames (quarterly,

annual), (5) distributed lag structures (1, 4, and 8 quarters), (6) output measures

(production, labor hours), (7) degrees of non-linearity in MSHA effects (linearity, cubic,

orthogonal polynomials), (8) time effects (yes, no), (9) location effects (yes, no), (10)

non-exogeneity of MSHA’s mine-specific abatement activities (yes, no), and (11)

estimation techniques (GMM, OLS, Tobit, Heckit, count models). The conclusion

emerging is that the results in the overwhelming number of cases are unfavorable to the

safety enhancement objective of MSHA at current levels of regulation.

Table 2 presents definitions of the regression variables. In using the Arellano-

Bond dynamic panel model (4.1) on our quarterly mining data the focal dependent

injury, including death. Exogenous variables include quarterly time dummies

and mine district location dummies, which are described in Appendix B. Always treated

as predetermined is our primary measure of mining activity, the log of total employee

We consider three specifications for MSHA activities: models with general (minedistrict

level) deterrence measures, models with specific (to the mine itself) deterrence

measures, and models with both general and specific deterrence measures. Here the

mine’s enforcement district average monetary penalty per inspection with monetary

penalty (calculated excluding mine i), pennum is the mine’s enforcement district’s

inspections per mine with monetary penalty (excluding mine i), and wonum is the mine’s

enforcement district’s inspections per mine with withdrawal order (excluding mine i).

inspections with monetary penalties, and sumwo is the mine’s number of inspections with

we examine specifications where m(general) is part of x and specifications where

substantial where the likelihood of an injury occurring is viewed to be likely, highly likely, or has already

occurred.

1 3

m(specific) is part of w and include up to four lagged values of both y and m on the righthand

side for symmetry in dynamic adjustment in y to past shocks and policy changes.

To fix ideas, the prototypical model specification we estimate is

4 4 4

1 1 2 2

1 0 0

= = =

in coal mine injuries here we are generally interested in equilibrium multiplier effects of

å

We could not produce a single regression using mine-specific abatement measures

that had an estimated negative effect for MSHA, which we attribute to the inability of the

instrumental variables approach to correct for the endogeneity of MSHA whereby

additional injuries in a mine trigger additional inspections. Regressions with specific

deterrence regressors that parallel our focal regression in terms of specification and

instrument sets appear in Appendix C. For our subsequent cost-effectiveness calculations

we selected the only regression from over 200 we estimated that simultaneously satisfied

the following criteria: computational feasibility (maximum lag length for an instrument is

15 quarters), quarterly data; time and location dummies; four-quarter lags on the injury

rate, production, and MSHA; at least one negatively signed MSHA coefficient that is

1.68 times its standard error; and the estimated equilibrium impact effect of MSHA is

The only (one-step Arellano and Bond) regression that satisfied the intersection of

our model selection criteria just described produced the following result, where underline

indicates that the coefficient was 1.68 times its (robust) standard error:

because of the large amount of data mining behind the regression result. A useful approximate result

described in Lovell (1983) for the connection between the true and claimed levels of significance is that

explanatory variables’ coefficients.

1 4

Our focal regression (6.2) yields an equilibrium impact multiplier for lindam that is

small, positive and statistically insignificant (which we will subsequently ignore in our

policy simulations), an equilibrium impact multiplier for penum that is –0.31, which

implies an elasticity at the means of – 0.25, and an equilibrium impact multiplier for

both of the estimated MSHA effects in (6.2) are close to the results in Scholtz and Gray

(1990) for the general deterrence effects of OSHA.

Before considering the economic and policy implications of our results we note

that some might view omitting possible health improvements from MSHA inspection

activities as a gap in our research. In 1970, the year after passage of the Coal Mine Act,

the number of death listings with any mention of coal workers’ pneumoconiosis (black

lung disease) was 2,189; by 1996 the number of death listings had dropped 35 percent to

1,417 (U.S. Department of Health and Human Services 1991, 1999). The incidence of

coal workers’ pneumoconiosis fell even more dramatically than the number of death

dynamic panel regression models, which are OLS, IV fixed effects and IV first-differences (Blundell,

Bond, and Windmeijer 2000; Bond 2002). In the IV fixed effects results no general deterrence coefficient

was at least 1.68 times its standard error, and the results from IV first differences did not satisfy the basic

MSHA are positive when we smooth our quarterly data by annualizing it.

1 5

listings. During the first round of the NIOSH Coal Workers’ X-Ray Surveillance

Program (1970–1973), 11 percent of miners had some form of coal workers’

pneumoconiosis. During the sixth round of surveillance (1992–1996), 2.8 percent of

miners had some form of coal workers’ pneumoconiosis, which is about a 75 percent

drop from the initial level (U.S. Department of Health and Human Services 1999).

Although MSHA may have been a factor in improving miner health, other factors may

also have contributed, such as improvements in technology, union efforts, greater worker

awareness, and even reductions in smoking incidence.

Attempting to disentangle all the potential influences on miner health would be

difficult econometrically, to say the least. Even more problematic would be trying to

relate health improvements to specific inspections given the long lag-time between

worker exposure and any signs of worker ill health. Our research, therefore, focuses on

the impact of MSHA general and safety-related inspections on miner safety. We exclude

from our empirical work enforcement activities not on mine property, including computer

generated dust sampling, education and training activities, and inspections geared

specifically toward health issues. Because general and safety-related inspections uncover

few health-related problems, changes in the number of the inspections we examine

We now turn our attention to the arithmetic details of safety inspections’ costs and

benefits. Viscusi (1993) and Viscusi and Aldy (2002) argue that the range of reasonable

value-of-life estimates is from $3 million to $7 million and that the value of a lost

workday injury is about $50,000 ($1990). The highest reported implicit value of injury in

Viscusi (1993) is Biddle and Zarkin’s (1988) estimate based on willingness to accept,

$131,495. We base our calculations of the costs and effectiveness of MSHA on the

estimates of the economic losses from fatal and non-fatal injuries just mentioned.

believed to result in injury). Of the almost half million violations in our data 94 percent were discovered

during either general or safety related inspections and six percent were found during other MSHA

enforcement activities. In the inspections in our estimation sample, over 99 percent of the citations were for

safety violations.

1 6

During 1983–1997 there were 428 fatalities and 91,773 nonfatal lost workday

injuries in our estimation sample. The proportion of fatal injuries in all injuries was

0.0046, and the corresponding proportion of nonfatal lost workday injuries in all injuries

was 0.9954. The value of reducing one injury established earlier in (2.1) is

where VL is the value of a life saved, and VI is the value of an injury prevented.

Using the highest value of life and value of injury figures mentioned above so as

to make the gains from MSHA as large as possible, the benefit of reducing an injury in

underground coal mines (converted to $2002) is

In Section 6 we located the one of approximately 200 regressions with the largest

significant estimated effects of wonum and pennum. Based on our cherry picked

regression the largest possible injury reducing effect of inspections leading to a

withdrawal order or a monetary penalty is then

( ) 2.53 0.306

¶ ¶

=- -

¶ ¶

Here,

´

´

calculations use a constant value for the impact of MSHA. The cubic in MSHA that will capture the nonlinearity

orthogonal polynomials regression. In the orthogonal polynomials regressions paralleling (6.1) the only

ineffectiveness of MSHA at relatively high or low levels of enforcement depicted in Figure 1. On the other

hand we also estimated simple dynamic censored (ZINB, Tobit, and Heckit) regression models that take

explicit account of the fact that about 50 percent of the observations on the dependent variable are zero.

The results show that MSHA has no effect at the extensive margin and all of its effect is at the intensive

1 7

From 1983 to 1997 for the mines in our estimation sample there were 6,249

inspections with at least one withdrawal order for a serious violation, 79,888 inspections

with a monetary penalty for at least one serious violation, and 252,411 inspections. The

fraction of withdrawal order inspections in all inspections was 0.0248, and the fraction of

monetary penalty inspections in all inspections was 0.3165. In the last quarter of 1997 the

number of mines that had been operating for at least 5 quarters (the minimum necessary

to be in the estimation sample) was 572.

Substituting the inspection and mines’ operating numbers into equations (7.4) and

(7.5) and then substituting the derivatives of equations (7.4) and (7.5) with respect to I

into (7.3) yields

0.0248 0.3165

( ) 2.55 0.306 0.00011 0.00017

572 572

inspections MSHA would need to eliminate one workplace injury. Using the upper bound

regression results from Section 6 produces a lower bound for dI = 3,584.

We have MSHA supplied information on inspector time for 99 percent of the

252,411 inspections of the underground coal mines in our estimation sample. Time is

broken down into four categories: travel, report writing, surface inspections, and

mechanized mining unit inspections. For the total sample the average and median total

inspection times were 29.8 hours and 8 hours. Excluding the longest 1 percent of

inspections by total time, the average and median inspection lengths were 24.7 hours and

8 hours.

According to the Position Classification Standard for Mine Safety and Health,

GS-1822, a starting underground coal mine inspector would have a government service

classification of 9. In 2001, GS-9, step 1 received an hourly wage of $15.93

(http://www.opm.gov/oca/01tables/gs hrly/html/01gshr.htm).

Ignoring overhead costs and using the median inspection length, the minimum

cost of the 3,584 additional inspections needed to reduce total coal mine injuries by one

margin, which is imply that at low levels of injury MSHA is ineffective at reducing injuries (to zero). For

econometric background on sophisticated censored dynamic panel models see Hu (2002).

1 8

What then is the cost-benefit ratio when we ignore the fact that we cherry-picked

regression results to get the most favorable impact of MSHA and in turn use the least

possible cost of an inspection? The estimated cost of eliminating an injury is $463,966

and the benefit from eliminating an injury is $219,443. The implied cost/benefit ratio for

the most favorable case we can construct for MSHA is about 2.1 > 1. At current levels

safety inspections are not cost-beneficial.

We might also address the cost-effectiveness issue somewhat differently and

focus on either reducing fatalities in isolation or on reducing non- fatal injuries in

isolation. Because 0.46 percent of all injuries are fatalities, to eliminate a fatality one

would require reducing total injuries by 1/0.0046 = 217. Because few injuries in mining

are fatal, equation (6.6) indicates that the number of additional inspections required to

eliminate one miner death would then be 779,155. Evaluated at the median time per

$100,865,530.

As a reference point for comparison and evaluation we can consider that

regulatory allocations involve an opportunity cost as they impose real financial costs on

consumers and taxpayers because the money spent on regulatory costs would otherwise

be spent on other bundles of consumer commodities. Based on such risk-risk tradeoff

considerations, economists have estimated that when government agencies propose risk

reducing regulations that impose a cost per life saved at levels of about $69 million or

more ($2002), then on balance the regulation is harming individual health (Viscusi 1994,

1998). So, it is important to recognize that the MSHA cost of saving a life is about 1.5

times the cutoff point for an acceptable life-saving regulation from broad social

perspective that a policy analyst should use.

To put the amount of additional inspections needed to reduce fatalities by one into

perspective (again ignoring the general lack of statistical significance of MSHA safety

inspections), in 1997 the total coal enforcement budget for MSHA was $107 million

(Budget of the United States Government, Fiscal Year 1999). In 1997 there were 72,390

1 9

which includes more than just labor cost of the inspector. In $2002, the cost of

10 times the annual enforcement budget. Put differently, the increase in inspections

needed to eliminate one miner death is more than 10 times the total number of

inspections now conducted by MSHA.

To eliminate one non- fatal injury would require an additional 3,601 inspections

using equation (7.6). The lower bound estimate of eliminating a non-fatal injury using the

$466,167, which is over 2.6 times the estimated benefit of an injury foregone.

It has frequently been suggested that regulatory programs be subjected to

continued OMB review for cost and effectiveness (Kniesner and Viscusi 2003 and

references therein). We close with an example of how a cost-effectiveness review could

be applied to MSHA because it may help to frame the policy implications of our

empirical results. It will make things more transparent, too, to recast our estimated

MSHA effects in terms of life years gained, on balance, if some of the MSHA

enforcement budget were reallocated to a few identifiable programs likely to affect the

health of miners.

As we have noted, the proportion of fatal injuries to all injuries is 0.0046, and the

proportion of nonfatal lost workday injuries to all injuries is 0.9954. Totinj combines

both fatal and nonfatal lost workday injuries. The number of lost life years saved from

reducing one injury is then

The average age of miners killed on the job in the estimation sample is 37.9.

Based on life expectancy tables posted at the National Center for Health Statistics the

2 0

remaining life expectancy of a 38 year old is 40.7 years. Using a 5 percent real interest

rate (as applied in Tengs et al. 1995) the discounted number of life years saved from

avoiding one mining death is then 17.3.

The average days lost from work due to a nonfatal injury in the estimation sample

is 39.17. We calculate average days lost replacing reported days lost with statutory days

lost for all permanent total or permanent partial disabilities with reported days lost of

zero. On average, a miner loses 0.107 of a work year from a nonfatal injury. Substituting

17.3 for lost years due to death and 0.107 for lost years due to nonfatal injury into the

equation above, the number of life years saved for every miner injury avoided is 0.186.

From previous calculations, the minimum cost of avoiding one injury using

government inspectors’ salary rates expressed in $2002 is $463,996. Combining results,

Our estimates imply that using the most optimistic estimated effects from Section

7, it costs about $2,500,000 per life year gained via MSHA enforcement activities. The

appropriate public policy issue, then, is whether there are cheaper ways to improve the

health of the population overall or of miners in particular.

A rich source of information for our ultimate policy evaluation exercise is again

Tengs et al. (1995), who calculate government cost per life year gained for 500 health

enhancing interventions. If one takes a transcendental view that a life year is a life year

no matter whose it is, then there are many programs Tengs et al. discover that have a per

consideration that any movement of resources out of MSHA’s safety enforcement budget

be put into programs likely to affect persons with the demographic characteristics of the

and investigations, enforcement activities not on mine property, and education and training evaluations.

reducing one fatality is at least $375,471,940 in 2002 dollars. Dividing by 17.3 (the number of discounted

life years), the cost per life year saved by reducing only mortality risk is then about $21,703,580.

requiring smoke detectors in homes, mandatory motorcycle helmet laws, banning residential growth in

tsunami -prone areas, banning sale of three-wheel ATVs, rubella vaccinations for children age two, and

smoking cessation advice for pregnant women who smoke. For the interested reader we note that the most

expensive programs per life year gained ($2002) include sickle cell screening for non-black low risk

newborns ($42 billion) and applying chloroform limits on private wells to emissions at the 48 worst case

pulp mills ($123 billion).

2 1

typical miner. What net gain in health, as measured by life years, could be obtained by

moving 25 percent of MSHA’s enforcement budget (having one fewer inspection per

mine per year) into alternative programs that could benefit the population of miners? The

results are surprising despite the relative small budgetary level of MSHA.

MSHA is small relative to other well-known regulatory agencies. In recent years

the overall budget of OSHA has been 1.7 times the budget of MSHA, and the annual

budget of the EPA has been 24.6 times the budget of MSHA. Let us just consider now the

annual enforcement budget of MSHA, whic h is about $110 million. One- fourth of the

MSHA enforcement budget is $27.5 million. At a cost of $2,500,000 per life year, a 25

percent reduction in MSHA inspections would reduce life years by about 11, which is

less than one statistical miner’s life. Using the list in Tengs et al. (1995), programs that

could affect the health of persons who might be in the population of miners would

include more heart disease screening or more on-site defibrillators, as suggested recently

by OMB, which would each produce a life year at a cost of $40. So, moving $27.5

million from the MSHA enforcement budget into more heart disease screening or

defibrillators would gain on balance 687,489 = (687,500 – 11) life years for the affected

population, which is equivalent to about 16,800 statistical miners’ lives.

The point of the exercise is to demonstrate that even a program as small as MSHA

can have relatively large opportunity costs. We have shown that a modest amount of

reallocation of program expenditures can make a substantial improvement in the public

health of the target population. Although there are specific mandates via OSHA and

MSHA addressing workplace health and safety, funding levels for OSHA and MSHA and

their target activities are legislative decisions. As policy analysts, we argue for costeffective

government policy in the area of promoting health and longevity. Our estimates

demonstrate the sizeable potential gain in miners’ health from budgetary reallocation to

other existing programs. We believe our estimates clearly imply a need for government to

take a more transcendental view by considering public health more generally and

consider more comprehensively the options available to improve the health of the

working population.

2 2

All 32.2 167.2 0 1 6 112 4,656 36,704

Failure to abate 104B 59.4 221.5 0 3 25 282 4,633 8,651

Unwarrantable

failure to comply 104D1 22.9 121.8 0 1 5 52 3,431 8,186

Subsequent similar 104D2 15.3 125.8 0 0 2 26 1,777 13,781

Imminent danger 107A 43.9 202.6 0 1 8 178 4,656 6,086

Source: Authors’ calculations.

2 3

The enforcement of MSHA standards is divided between the Coal Mine Safety

and Health and Metal and Nonmetal Mine Safety and Health groups. In turn the groups

are broken down into enforcement districts (11 in coal and 6 metal and nonmetal) and

field offices (65 in coal and 50 metal and nonmetal). The 11 coal mining enforcement

districts are:

District 1 Anthracite coal mining regions in Pennsylvania

District 2 Bituminous coal mining regions in Pennsylvania

District 3 Maryland, Ohio, and Northern West Virginia

District 4 Southern West Virginia

District 5 Virginia

District 6 Eastern Kentucky

District 7 Central Kentucky, North Carolina, South Carolina, and Tennessee

District 8 Illinois, Indiana, Iowa, Michigan, Minnesota, Northern Missouri, and

Wisconsin

District 9 All states west of the Mississippi River, except Minnesota, Iowa, and

Northern Missouri

District 10 Western Kentucky

District 11 Alabama, Georgia, Florida, Mississippi, Puerto Rico, and Virgin Islands

Besides conducting inspections, the regional offices also review mine plans for

safety concerns. The mine operator devises appropriate engineering plans and then the

engineering specialists at MSHA review and approve the proposed plans. Once approved,

the mine operator must follow the plans. Specific areas include control of mine roof and

ventilation system.

District managers are responsible for supervising inspectors in their districts.

MSHA has acknowledged that there is inconsistency in how inspectors interpret

standards. To help remedy the problem, it has established a District Managers Council

(DMC) which meets quarterly to discuss and try to correct enforcement inconsistencies.

The Office of Assessments determines the size of monetary penalties. The criteria

for penalties include the size of the business, the seriousness of the violation, and the

degree of the mine operator’s negligence. When a major accident is reported, the district

manager dispatches MSHA personnel to the site. The mine operator has control and

responsibility for rescue efforts but must seek approval from MSHA for actions taken.

2 4

In a report by the Office of the Inspector General on metal/nonmetal mining

enforcement and compliance assistance activities it was recommended that MSHA should

improve guidance to district offices regarding program implementation and operation to

enhance consistency in program performance and management. The report also found

disparities in the inspector resources available per mine on a district basis. Factors that

should be considered in allocating inspector resources include mine size, geographic

clustering, and travel time. The report also discovered that the mix of activities between

enforcement and compliance assistance fluctuated among the districts and within a

district from year to year. There was no consensus among district managers about the

relative effectiveness of enforcement activities and compliance-oriented activities. All of

the district managers believed both types of activities had merit but the difficulty was

allocating time between activities. (Mine Safety and Health Administration 2001.)

25

2 6

Arellano, Manuel and Stephen Bond, “Some Tests of Specification for Panel Data: Monte

Carlo Evidence and an Application to Employment Equations,” Review of Economic

Studies, 58:2 (April 1991), pp. 277–297.

Auld, M. Christopher, Emery, J.C. Herbert, Gordon, Daniel V., and Douglas McClintock,

“The Efficacy of Construction Site Safety Inspections,” Journal of Labor Economics,

19:4 (October 2001), pp. 900–921.

Biddle, Jeff E. and Gary Zarkin, “Worker Preferences and Market Compensation for Job

Risk,” Review of Economics and Statistics, 70:4 (November 1988), pp. 660–667.

Blundell, Richard, Bond, Stephen, and Frank Windmeijer, “Estimation in Dynamic Panel

Data Models: Improving on the Performance of the Standard GMM Estimators,” London:

Institute for Fiscal Studies, Working Paper WP00/12.

Boal, William M., “The Effect of Unionism on Accidents in Coal Mining, 1897–1929,”

Des Moines, IA: College of Business and Public Administration, Drake University,

Working Paper, November 2003.

Bond, Stephen, “Dynamic Panel Data Models: A Guide to Micro Data Methods and

Practice,” London: Institute for Fiscal Studies, Centre for Microdata Methods and

Practice (CEMMAP), Working Paper CW09/02.

Fishback, Price, Soft Coal, Hard Choices: The Economic Welfare of Bituminous Coal

Miners, 1890–1930, New York: Oxford University Press, 1992.

Fuess, Scott M. and Mark A. Lowenstein, “Further Analysis of the Theory of Economic

Regulation: The Case of the 1969 Coal Mine Health and Safety Act,” Economic Inquiry,

XXVIII (April 1990), pp. 354–389.

Greene, William H., Econometric Analysis, Fifth Edition, Upper Saddle River, NJ:

Prentice Hall, 2003.

Hahn, Robert W. and Robert E. Litan, “An Analysis of the Fifth Government Report On

the Costs and Benefits of Federal Regulations,” Washington, DC: AEI-Brookings Joint

Center for Regulatory Studies, Regulatory Analysis 02–2, May 2002.

Hahn, Robert W. and Cass R. Sunstein, “A New Executive Order for Improving Federal

Regulation? Deeper and Wider Cost-Benefit Analysis,” Washington, DC: AEI-Brookings

Joint Center for Regulatory Studies, Working Paper 02–4, March 2002.

Hall, Peter and Joel L. Horowitz, “Bootstrap Critical Values for Tests Based on

Generalized-Method-of-Moments Estimators,” Econometrica, 64:4 (July 1996), pp. 891–

916.

2 7

Hu, Luojia, “Estimation of a Censored Dynamic Panel Data Model,” Econometrica, 70:6

(November 20002), pp. 2499–2517.

Kniesner, Thomas J. and John D. Leeth, Simulating Workplace Safety Policy, Boston:

Kluwer Academic Publishers, 1995.

Kniesner, Thomas J. and W. Kip Viscusi, “Why Relative Position Does Not Matter: A

Cost-Benefit Analysis,” Yale Journal on Regulation, 20:1 (Winter 2003), pp. 1–24.

Lewis-Beck, Michael S. and John R. Alford, “Can Government Regulate Safety? The

Coal Mine Example,” American Political Science Review, 74:3 (September 1980), pp.

745–756.

Lovell, Michael C., “Data Mining,” Review of Economics and Statistics, 65:1 (February

1983), pp. 1–12.

Lutter, Randall, Morrall III, John F., and W. Kip Viscusi, “The Cost-Per-Life-Saved

Cutoff for Safety-Enhancing Regulations,” Economic Inquiry, 37:4 (October 1999), pp.

599–608.

Neumann, George R. and Jon P. Nelson, “Safety Regulation and Firm Size: Effects of the

Coal Mine Health and Safety Act of 1969,” Journal of Law and Economics, XXV

(October 1982), pp. 183–199.

Office of Management and Budget, Executive Office of the President of the United

States, Budget of the United States Government, Fiscal Year 1999, Washington, DC:

Government Printing Office.

Sala-I-Martin, Xavier X., “I Just Ran Two Million Regressions,” American Economic

American Economic Association, 87:2 (May 1997), pp. 178–183.

Scholtz, John T. and Wayne B. Gray, “OSHA Enforcement and Workplace Injuries: A

Behavioral Approach to Risk Assessment,” Journal of Risk and Uncertainty, 3:3

(September 1990), pp. 283–305.

Sunstein, Cass R., Risk and Reason: Safety, Law, and the Environment, Cambridge, UK:

Cambridge University Press, 2002a.

Sunstein, Cass R., The Cost-Benefit State: The Future of Regulatory Policy, Chicago:

American Bar Association, 2002b.

Tengs, Tammy O., Adams, Miriam E., Pliskin, Joseph S., Safran, Dana Gelb, Siegel,

Joanna E., Weinstein, Milton D., and John D. Graham, “Five-Hundred Life-Saving

Interventions and Their Cost-Effectiveness,” Risk Analysis, 15:3 (June 1995),

2 8

pp. 369–390.

Tengs, Tammy O. and John D. Graham, “The Opportunity Costs of Haphazard Social

Investments,” in Robert W. Hahn (editor), Risks, Costs, and Lives Saved: Getting Better

Results from Regulation, New York: Oxford University Press, 1996, pp. 167–182.

U.S. Department of Health and Human Services, Work-Related Lung Disease

Surveillance Report 1991, Washington, DC: U.S. Department of Health and Human

Services, NIOSH Publication No. 91-113, October 1991.

U.S. Department of Health and Human Services, Work-Related Lung Disease

Surveillance Report 1999, Washington, DC: U.S. Department of Health and Human

Services, NIOSH Publication No. 2000-105, December 1999.

U.S. Department of Labor, Study of the Metal/Nonmetal Mining Enforcement and

Compliance Assistance Activities, 1983–2000, Washington, DC: Mine Safety and Health

Administration, Report No. 2E-06-620-003, September 26, 2001.

Viscusi, W. Kip, Fatal Tradeoffs: Public and Private Responsibilities for Risk, New

York: Oxford University Press, 1992, pp. 181–205.

Viscusi, W. Kip, The Value of Risks to Life and Health,” Journal of Economic

Literature, 31:4 (December 1993), pp. 1912–1946.

Viscusi, W. Kip, “Mortality Effects of Regulatory Costs and Policy Evaluation Criteria,”

Rand Journal of Economics, 25:1 (Spring 1994), pp. 94–109.

Viscusi, W. Kip, Rational