Alison Morantz, Coal Mine Safety: Do Unions Make a Difference? 66 Indus. & Lab. Rel. Rev. 88 (2013).
Note on data availability: the data used in this paper was obtained from government agencies including Mine Safety and Health Administration (MSHA), Energy Information Administration (EIA), and National Institute for Occupational Safety and Health (NIOSH).
The mine union status data, which is critical to the paper, was provided by EIA. Unfortunately, EIA does not allow the author to distribute
this data to third parties, which makes sharing the data and programs for replication impossible. If you would like to obtain the data used
for this paper, please contact Joseph Conklin at joseph dot conklin at eia dot gov if the date is before August 30, 2013, or Jacob Bournazian
at jacob dot bournazian at eia dot gov after August 30, 2013.
Although the United Mine Workers of America (UMWA) has always advocated strongly for miners’ safety, the empirical literature contains no evidence that unionization reduced mine injuries or fatalities during the 1970s and ‘80s. The author uses an updated methodology and a more comprehensive data set than previous studies to examine the relationship between unionization and underground, bituminous coal mine safety from 1993 to 2010. She finds that unionization predicts a substantial and statistically significant decline in traumatic injuries and fatalities, the two safety measures that are the least prone to reporting bias. These results are especially pronounced among larger mines. Overall, unionization is associated with a 14 to 32% drop in traumatic injuries and a 29 to 83% drop in fatalities. Yet unionization also predicts higher total and nontraumatic injuries, suggesting that injury reporting practices differ between union and nonunion mines.
Note: Published versions only report coefficients for variables of interest. "Expanded" tables report coefficients for all covariates except for certain dummy variables (e.g. quarter dummies). "Comprehensive" tables report all coefficients for all covariates.
County information was provided by MSHA. The county-level mine counts displayed in the figure include all 2,645 underground bituminous coal mines that were active for at least one quarter between 1993 and 2010. (Note that ten mines depicted in this figure were missing information on controller, without which controller size could not be calculated, and therefore were excluded from the final sample of 2,635 mines used in the regression analysis.) Due to high rates of entry and exit in the industry, no more than half of the sample was active in any given quarter.
For each year from 1993 to 2010, the figure displays the percentage of all underground, bituminous coal mines with non-zero production and hours worked that were unionized.
For each year from 1993 to 2010, the figure displays, respectively, total injuries and traumatic injuries per 2,000 hours worked for underground, bituminous coal mines. Figures are reported separately for unionized and non-unionized mines. For each given year and mine type, injuries and hours are summed across all mines.
Table 01a. Effect of Union Status on Non-Traumatic Injuries in Underground Subunit per Mine-Quarter
Table 2. Effect of Union Status on Injury Frequency: Baseline Models
Specification/Version:
FTE Pub.
FTE Priv.
Employees Pub.
Employees Priv.
Tonnage Pub.
Tonnage Priv.
Mine/Controller Size Units:
100 FTEs
100 FTEs
100 Employees
100 Employees
Millions of Tons
Millions of Tons
Nontraumatic Injury Model
1.351***
1.258***
1.376***
1.290***
1.363***
1.288***
(0.07)
(0.10)
(0.07)
(0.10)
(0.07)
(0.09)
Total Injury Model
1.148***
1.045
1.163***
1.062
1.141***
1.043
(0.05)
(0.07)
(0.05)
(0.07)
(0.05)
(0.06)
Traumatic Injury Model
0.765***
0.687***
0.767***
0.687***
0.762***
0.683***
(0.04)
(0.05)
(0.04)
(0.05)
(0.04)
(0.04)
Observations
38,890
24,593
38,890
24,593
38,890
24,593
# of Union Mines / # of Total Mines
355 / 2,635
186 / 1,684
355 / 2,635
186 / 1,684
355 / 2,635
186 / 1,684
Fatality Model
0.346***
0.421*
0.358***
0.436*
0.369***
0.423**
(0.13)
(0.19)
(0.13)
(0.20)
(0.13)
(0.18)
Observations
11,045
6,948
11,045
6,948
11,045
6,948
# of Union Mines / # of Total Mines
352 / 2,568
182 / 1,644
352 / 2,568
182 / 1,644
352 / 2,568
182 / 1,644
Significance Levels: *** 1%, ** 5%, * 10%. Standard errors, clustered at the mine level, are shown in parentheses.
Results Presented: The table reports IRR (incidence rate ratio)a coefficients on the union indicator variables in negative binomial regression models. Hours worked is the exposure term.
Definitions: A quarterly FTE is defined as 500 hours worked.
Unit of Observation: The unit of observation is the mine-quarter for the nontraumatic, total, and traumatic injuries regressions. The unit of observation is the mine-year for fatality regressions.
Dependent Variables: The dependent variables are counts of injuries of each type (specified in the far-left column) that occur underground. Traumatic injuries are defined to include the following: amputations; enucleations; fractures; chips; dislocations; foreign bodies in eyes; cuts and lacerations; punctures; burns/scalds; crushings; chemical, electrical, and laser burns; and fatalities. (See footnote 13 for more details on the definition on traumatic injuries.) The sum of traumatic and nontraumatic injuries comprises total injuries.
Independent Variables: All models include the following regressors: union dummy, mine size (a continuous variable whose units are specified in column header), union X mine size, logged controller size (a continuous variable whose units are specified in column header), mine age, mine productivity, total lost-work injuries (in hundreds) during previous calendar year (for fatality models) or previous four quarters (for non-fatality models), total penalty points (in thousands) during previous calendar year (for fatality models) or previous four quarters (for non-fatality models), dummies indicating presence of each respective mine subunit, quarter/year dummies, district dummies, and a constant term. Public-fields versions also include a longwall indicator. Confidential-fields versions also include the number of coal beds, mean coal bed thickness (in yards), subsidiary indicator, captive production as a percentage of total production, recoverable coal reserves, and mining method percentages. See Appendix C for complete variable definitions. An expanded version of this table, including a full covariate report, is available at http://amorantz.stanford.edu/companions/union-coal-mine-safety/.
Sample: The sample consists of underground bituminous coal mines with positive coal production and positive hours worked. The public-fields versions contain mine-quarters from 1993–2010, whereas the confidential-fields versions are restricted to 1998–2010. Because the historical variables (lost-work injuries and penalty points) are summed up over the previous four quarters in the nontraumatic, total, and traumatic injuries regressions but are summed up over the previous calendar year in the fatality regressions, some mines excluded from the fatality models are included in the other models. For example, if a mine is open for all of only one calendar year, it will have no historical data at the yearly level, but it will have historical data for three of the four quarters it was open. The number of union mines is computed by counting the mines that were unionized for any of the mine-quarters in the sample period. The number of total mines is computed by counting each mine in the sample, regardless of union status. Note that the confidential-fields versions of all models exclude any observations that lack information on confidential covariates.
aA coefficient of 1 indicates no change at all in predicted injuries; coefficients between 0 and 1 represent a predicted fall in injuries (e.g. a coefficient of 0.97 represents a 3% decline); and coefficients greater than one represent predicted increases (e.g. a coefficient of 1.03 represents a 3% rise).
Table 3. Effect of Union Status on Injury Frequency: Discrete Size Groups
Table 3. Effect of Union Status on Injury Frequency: Discrete Size Groups
Specification/Version:
Nontraumatic Injuries
Total Injuries
Traumatic Injuries
Fatalities
Union X Mine Size Quartile 1
1.223
1.227*
1.221
0.000***
(0.16)
(0.14)
(0.19)
(0.00)
Union X Mine Size Quartile 2
1.237***
1.170**
0.987
0.321
(0.10)
(0.08)
(0.09)
(0.32)
Union X Size Mine Quartile 3
1.424***
1.195***
0.815***
0.824
(0.12)
(0.08)
(0.05)
(0.51)
Union X Mine Size Quartile 4
1.249***
1.000
0.697***
0.307***
(0.07)
(0.05)
(0.04)
(0.09)
Observations
38,890
38,890
38,890
11,045
# of Union Mines / # of Total Mines
355 / 2,635
355 / 2,635
355 / 2,635
352 / 2,568
Significance Levels: *** 1%, ** 5%, * 10%. Standard errors, clustered at the mine level, are shown in parentheses.
Results Presented: The table reports IRR (incidence rate ratio)a coefficients on the union indicator variables in negative binomial regression models. Hours worked is the exposure term.
Definitions: A quarterly FTE is defined as 500 hours worked.
Unit of Observation: The unit of observation is the mine-quarter for the nontraumatic, total, and traumatic injuries regressions. The unit of observation is the mine-year for fatality regressions.
Dependent Variables: The dependent variables are counts of injuries of each type (specified in the top row) that occur underground. Traumatic injuries are defined to include the following: amputations; enucleations; fractures; chips; dislocations; foreign bodies in eyes; cuts and lacerations; punctures; burns/scalds; crushings; chemical, electrical, and laser burns; and fatalities. (See footnote 13 for more details on the definition on traumatic injuries.) The sum of traumatic and nontraumatic injuries comprises total injuries.
Independent Variables: All specifications presented above rely exclusively on regressors that are publicly available. In addition to discrete union-size interaction terms, all models include the following regressors: size quartiles (as determined by total FTEs), logged controller size (a continuous variable reflecting the controller’s total number of FTEs), mine age, mine productivity, total lost-work injuries (in hundreds) during previous calendar year (for the fatality model) or previous four quarters (for non-fatality models), total penalty points (in thousands) during previous calendar year (for the fatality model) or previous four quarters (for non-fatality models), dummies indicating presence of each respective mine subunit, quarter/year dummies, district dummies, longwall indictor, and a constant term. See Appendix C for complete variable definitions. An expanded version of this table, including a full covariate report, is available at http://amorantz.stanford.edu/companions/union-coal-mine-safety/.
Sample: The sample consists of underground bituminous coal mines with positive coal production and positive hours worked from 1993–2010. Because the historical variables (lost-work injuries and penalty points) are summed up over the previous four quarters in the nontraumatic, total, and traumatic injuries regressions but are summed up over the previous calendar year in the fatality regressions, some mines excluded from the fatality models are included in the other models. For example, if a mine is open for all of only one calendar year, it will have no historical data at the yearly level, but it will have historical data for three of the four quarters it was open. The number of union mines is computed by counting the mines that were unionized for any of the mine-quarters in the sample period. The total number of mines is computed by counting each mine in the sample, regardless of union status. Note that the confidential-fields versions of all models exclude any observations that lack information on confidential covariates.
a A coefficient of 1 indicates no change at all in predicted injuries; coefficients between 0 and 1 represent a predicted fall in injuries (e.g. a coefficient of 0.97 represents a 3% decline); and coefficients greater than one represent predicted increases (e.g. a coefficient of 1.03 represents a 3% rise).
Table 4. Effect of Union Status on Nontraumatic Injuries in Underground Subunit per Mine-Quarter
Table 4. Effect of Union Status on Nontraumatic Injuries in Underground Subunit per Mine-Quarter
Specification/Version:
FTE Pub.
1993-1998
1999-2004
2005-2010
Non-Traumatic Injury Model
1.351***
1.497***
1.245**
1.268*
(0.07)
(0.09)
(0.12)
(0.16)
Total Injury Model
1.148***
1.316***
1.050
0.988
(0.05)
(0.07)
(0.09)
(0.10)
Traumatic Injury Model
0.765***
0.921
0.669***
0.686***
(0.04)
(0.06)
(0.06)
(0.08)
Observations
38,890
16,629
11,460
10,801
# of Union Mines / # of Total Mines
355 / 2,635
294 / 1,765
129 / 1,141
65 / 928
Fatality Model
0.346***
0.378*
0.331*
0.552
(0.13)
(0.19)
(0.20)
(0.38)
Observations
11,045
4,763
3,308
2,974
# of Union Mines / # of Total Mines
352 / 2,568
290 / 1,690
128 / 1,093
65 / 903
Significance levels: *** 1%, ** 5%, * 10%. Standard errors, clustered at the mine level, are shown in parentheses.
Results Presented: The table reports IRR (incidence rate ratio)a coefficients on the union indicator variables in negative binomial regression models. Hours worked is the exposure term. The "FTE Public (Baseline)" column contains coefficient estimates from the principal baseline models (using 100 quarterly FTEs as the size measure and relying exclusively on public data) presented in Table 2. The results presented in the other three columns correspond, respectively, to coefficient estimates from identical models run on six-year subsamples.
Definitions: A quarterly FTE is defined as 500 hours worked.
Unit of Observation: The unit of observation is the mine-quarter for the nontraumatic, total, and traumatic injuries regressions. The unit of observation is the mine-year for fatality regressions.
Dependent Variables: The dependent variables are counts of injuries of each type (specified in the far-left column) that occur underground. Traumatic injuries are defined to include the following: amputations; enucleations; fractures; chips; dislocations; foreign bodies in eyes; cuts and lacerations; punctures; burns/scalds; crushings; chemical, electrical, and laser burns; and fatalities. (See footnote 13 for more details on the definition on traumatic injuries.) The sum of traumatic and nontraumatic injuries comprises total injuries.
Independent Variables: All specifications presented above rely exclusively on regressors that are publicly available. All models include the following regressors: union dummy, mine size (a continuous variable reflecting the mine’s total number of FTEs) , union X mine size, logged controller size (a continuous variable reflecting the controller’s total number of FTEs), mine age, mine productivity, total lost-work injuries (in hundreds) during previous calendar year (for fatality models) or previous four quarters (for non-fatality models), total penalty points (in thousands) during previous calendar year (for fatality models) or previous four quarters (for non-fatality models), dummies indicating presence of each respective mine subunit, quarter/year dummies, district dummies, a longwall indicator, and a constant term. See Appendix C for complete variable definitions. An expanded version of this table, including a full covariate report, is available at http://amorantz.stanford.edu/companions/union-coal-mine-safety/.
Sample: The sample consists of underground bituminous coal mines with positive coal production and positive hours worked from 1993–2010. Because the historical variables (lost-work injuries and penalty points) are summed up over the previous four quarters in the non-traumatic, total, and traumatic injuries regressions but are summed up over the previous calendar year in the fatality regressions, some mines excluded from the fatality models are included in the other models. For example, if a mine is open for all of only one calendar year, it will have no historical data at the yearly level, but it will have historical data for three of the four quarters it was open. The number of union mines is computed by counting the mines that were unionized for any of the mine-quarters in the sample period. The total number of mines is computed by counting each mine in the sample, regardless of union status. Note that the confidential-fields versions of all models exclude any observations that lack information on confidential covariates.
a A coefficient of 1 indicates no change at all in predicted injuries; coefficients between 0 and 1 represent a predicted fall in injuries (e.g. a coefficient of 0.97 represents a 3% decline); and coefficients greater than one represent predicted increases (e.g. a coefficient of 1.03 represents a 3% rise).
