Political Economy of Monitoring and Enforcement in the Coal Mining - - PowerPoint PPT Presentation

Political Economy of Monitoring and Enforcement in the Coal Mining Industry

R.J. Briggs, Anastasia Shcherbakova, Suman Gautum Pennsylvania State University 30 August 2012

Motivation

• Upper Big Branch Mine (Massey Energy) – West Virginia
• Coal dust explosion at 15:27
• 29 of 31 miners on site were killed
• Worst accident in U.S. since 1970
• MSHA found that safety violations contributed to explosion
• Mine permanently closed April 2012

Motivation

• 2009 – Upper Big Branch mine received 515 citations for

safety violations

• U.S. Mine Safety and Health Administration (MSHA) was

blamed for failing to enforce safety measures

• March 2012: former superintendent, Gary May, pleaded guilty to

impeding MSHA’s enforcement efforts

• Investigation reported that Massey Energy had a lot of

political power in the state

• FBI launched investigation of criminal negligence and possible

bribery of federal regulators

Motivation

• Political connections were a factor in the disaster
• Massey Energy contributed over $307,000 to federal political

candidates since the 1990 election cycle

• Former CEO Don Blankenship contributed tens of thousands

more

• The mining industry as a whole tripled its lobbying expenditures

from $10.2 million in 2004 to $30.8 million in 2008

• With respect to the effect of political activity on monitoring

and enforcement, was Massey Energy the exception or the rule?

Research Question

• What effect, if any, does lobbying have on coal mine violation

and fine outcomes?

Link Between Lobbying and Profits

• Firms maximize profits. Costly regulations may reduce profit

potential by increasing cost of doing business. If firms can successfully lobby against such regulations, they will increase profits.

• Lobbying represents firm’s efforts to fight costly regulations

(Jaffe & Palmer, 1995)

• Political expenditures signal a firm’s willingness to fight

against its political disinterests. As a consequence, large donors are inspected less and violate more (Gordon & Hafer, 2005)

• Firms lobby to raise an agency’s cost of performing its

functions

Literature Review

• Most prior research finds that political connections have little

value in developed economies (Jones and Olken 2005; Fisman, Fisman, Galef, and Khurana 2006)

• But corporate lobbying can affect firm outcomes (Drope and

Hansen 2006, Brasher and Lowery 2006, Kim 2008)

• Corporate spending on lobbying efforts exceeds direct or PAC

contributions to campaigns (Milyo et al. 2000)

• Lobbying can provide continuous rather than periodic influence,

regardless of which political party is in charge

• Lobbying is more prevalent in regulated industries; firms that

lobby “tend to outperform the market average and, to a lesser degree, the average peer in the same industry” (Kim, 2008)

Research Contribution

• Combine MSHA data on coal mine inspections and violations

with CRP data on lobbying expenditures by coal mine controllers

• To our knowledge, this study is:
• The first analysis of impacts of lobbying on regulatory outcomes
• The first empirical political economy examination of monitoring

and enforcement

• The first to separate inspections from violations in MSHA coal

mine data

Research Design

• Interested in four main outcomes:
• Probability of receiving a violation
• Number of violations received during a year
• Probability of securing a fine reduction
• Aggregate financial penalties accrued for all of year’s fines
• Main conditioning variables:
• Controller lobbying expenditures
• Mine inspections (examine only inspected firms)

Research Design

• Mine decides whether or not to comply with MSHA

regulations

• Decision based on which is greater, expected cost of

compliance or expected cost of violation

• Cost of compliance unobserved; we assume it exceeds

expected cost of violating for every instance in which a citation is issued

• therwise

] | [ ] 1 | [ if 1      viol violation C E violation C E viol

Research Design

• Expected cost of a single violation
• Since we only observe inspected mines,
• Total annual cost of violation:

E(fine) prob insp prob fine E detection prob viol C E       ) violation identifies ly successful insp ( ) ( ) ( ) ( ) 1 | (

1 ) (  insp prob

E(fine) prob #OwnInsp viol TC E     ) violation identifies ly successful insp ( ) 1 | (

Research Design

• Consider :
• According to MSHA guidelines, financial penalty assigned to

citation is based on:

1. History of previous violations 2. Size of operator’s business 3. Operator’s negligence 4. Gravity of violation 5. Operator’s good faith in trying to correct violation promptly 6. Effect of penalty on operator’s ability to stay in business

• Inspector allowed to propose alternative penalty when above

criteria results in “insufficient” fine

) ( fine E

Research Design

• Consider :
• Other factors to influence expected fine:
• Differences between regional inspection offices
• Individual inspector characteristics
• MSHA budgetary circumstances
• Etc.
• So likely to vary from inspection to inspection and

year to year, not directly observed by mine.

• Absorbed into coefficient on own inspections

) ( fine E ) ( fine E

Research Design

• Abstracting from improvements in detection mechanism

(except through increased inspection rates), and introducing lobbying activities, probability of choosing to violate regulations is

• First term picks up indirect monitoring and financial effects of

lobbying, second term accounts for remaining direct effect of lobbying on violation decisions

• And probability of violation will be a decreasing function of

the expected annual cost of violating )] 1 | ( [ ) 1 (    viol TC E f viol prob ) ( | ) 1 | ( lobbying f lobbying) E(fine #OwnInsp viol TC E    

Econometric model

1. Exploratory Granger causality tests on main variables of interest 2. Basic econometric specification:

Yi,t is one of four main outcomes of interest Lobbyc,t is controller‐level lobbying covariates Inspi,t is mine‐level inspection covariates Xi,t is mine‐level characteristics Zn,t is county‐level characteristics

t i t n t i t i t c t i

Z X Insp Lobby Y

, , 4 , 3 , 2 , 1 ,

           

Econometric model

• Two concerns:
• Regulator’s inspection activities may not be random (targeting)
• Decision to lobby may be endogenous (moral hazard)
• Instrumental variable approach:
• Inspections

1. Average annual inspection rate of each mine’s county peers 2. Distance from each mine to its regional inspection office

• Lobbying

1. Number of mines in controller’s portfolio

Econometric model

• Instrumenting for own inspections

1. Average annual inspection rate of each mine’s county peers (Shimshack & Ward 2005)

• Mine i’s violation behavior may affect its own inspection rate, but

will not affect inspection rates of all other mines in county.

• A rise in inspection rates of all other mines in county signals a

general increase in regulator activity and will be correlated with violation decision of mine i.

Econometric model

• Instrumenting for own inspections

2. Distance from each mine to its regional inspection office (Almeida & Carneiro 2008)

• Inspectors must visit a company’s facilities in person, so physical

distance between inspector and mine to be inspected is good proxy for regulator’s cost of inspection. Inspections will be less frequent where they are more costly.

• A mine’s location is determined exogenously by location of coal
• deposits. Assuming regulators don’t locate inspection offices

strategically, distance between mine and regulator is also determined exogenously.

• So distance will have no direct influence on a mine’s decision to

violate, but will affect frequency of inspector visits.

• Inspection frequency will in turn affect a mine’s compliance decision.

Econometric model

• Instrumenting for own lobbying

1. Number of mines in controller’s portfolio (Figueiredo & Silverman 2006)

• Studies point to positive correlation between a firm’s income and

profitability measures and engagement with lobbyists (more disposable cash gives a firm more resources to devote to lobbying activities).

• In our data, lobbying is done by mines’ controllers; we do not have

financial information on mine controllers, but know how many mines each oversees every year.

• Since lobbying is done by controllers, number of mines in controller’s

portfolio should not influence mine‐specific compliance decisions.

• But controllers in charge of many mines face greater potential

benefits of favorable industry laws, and therefore greater incentives to engage in lobbying.

Summary Statistics

Variable Mean Median Std dev Min Max N Number of inspections 12.8 6.0 24.2 1 190 3738 Number of violations 44.8 10.0 114.3 1609 3738 Number of S&S violations 15.1 3.0 37.7 445 3738 Number of non‐S&S violations 29.7 6.0 79.6 1242 3738 Final fines due (‘000 USD) 25.2 1.0 136.3 3993 3738 Final S&S fines (‘000 USD) 19.7 0.51 109.2 3067 3738 Final non‐S&S fines (‘000 USD) 5.5 0.4 29.3 926 3738

Granger Causality Tests

Y X F Inspections Violations 38.81*** Fines 15.43*** Lobbying 5.80** Violations Inspections 151.10*** Fines 33.05*** Lobbying 0.93 Fines Inspections 66.93*** Violations 105.34*** Lobbying 10.79*** Lobbying Inspections 6.98*** Violations 19.79*** Fines 77.31***

Significance levels: 10% * 5% ** 1% ***

All IV Results

Prob(violation) Number of violations Total Fines Prob(fine reduction) flag_lobby_xb 0.549088* 173.993370** 1.472861 0.305657 (0.320290) (84.816567) (0.902665) (0.255810) flag_lobby_lag1_xb ‐0.724790** ‐161.698255* ‐1.295054 ‐0.353706 (0.326850) (96.963574) (0.877708) (0.231784) log_lobby_xb 0.579688*** 27.358291 0.931491* 0.200601 (0.206738) (50.519284) (0.508593) (0.127275) log_lobby_lag1_xb ‐0.654316** ‐11.94469 ‐0.766997 ‐0.14382 (0.295087) (68.746124) (0.713592) (0.184240) log_lobby_others ‐0.309557 ‐13.252337 0.091072 0.251788 (0.486417) (122.983332) (1.200469) (0.343742) log_lobby_others_lag1 ‐1.61498 ‐51.625918 ‐1.622754 0.19781 (1.370244) (331.705564) (3.273834) (0.940742) insp_xb 0.000907 1.71025 0.028777 0.006931 (0.008493) (2.372041) (0.023012) (0.007415) insp_lag1_xb 0.008463 ‐4.044032* ‐0.046185*** ‐0.009778 (0.006935) (2.078714) (0.013972) (0.006483) Mineract 0.747411 37.002494 0.967306 ‐0.014506 (0.606256) (149.982793) (1.464945) (0.417937) IMR3 ‐168.229488*** ‐2.402268*** (46.147970) (0.391098) r2_a/r2_p 0.303 0.421 0.628 0.221 N 1529 1529 1529 1529 Note: Standard errors in parentheses; Significance levels: 10% * 5% ** 1% ***

Discussion

• Some evidence of lagged effect of lobbying on safety and

health outcomes

• Both decision to engage in lobbying expenditures and size of

donation appears to matter

• Lobbying may or may not be a public good
• Evidence of regulator targeting/effectiveness through higher

inspection rates

Thank you

Please send questions and comments to anastasia@eme.psu.edu

Probability of Violation

(1) (2) (3) (4) (5) (6) flag_controller_lobby 0.080140* 0.042862 0.097154** 0.051098 0.765927 0.549088* (0.041672) (0.043755) (0.048535) (0.049471) (1.215110) (0.320290) flag_controller_lobby_l1 0.106850*** 0.095784** ‐0.018858 ‐0.042301 ‐0.38524 ‐0.724790** (0.041300) (0.044414) (0.053835) (0.048464) (1.127996) (0.326850) log_lobby 0.034341* 0.040481* 0.181627*** 0.181620*** 1.306694*** 0.579688*** (0.018878) (0.021261) (0.050813) (0.048516) (0.466514) (0.206738) log_lobby_l1 ‐0.055786*** ‐0.056788*** ‐0.204945*** ‐0.192893*** ‐1.446037*** ‐0.654316** (0.019032) (0.021102) (0.048892) (0.045851) (0.487908) (0.295087) log_lobby_others 0.003736 0.062137*** 0.053159** 0.099898*** 1.048627*** ‐0.309557 (0.020211) (0.019597) (0.023576) (0.037256) (0.395151) (0.486417) log_lobby_others_l1 0.051240** 0.001777 0.032245 0.162455** 2.013531** ‐1.61498 (0.024343) (0.022555) (0.032515) (0.081136) (0.915621) (1.370244) Mineract ‐0.011372 0.055669** ‐0.00093 ‐0.051287 ‐0.802592* 0.747411 (0.021567) (0.021688) (0.024214) (0.040141) (0.427795) (0.606256) Insp 0.018978*** 0.012379*** 0.013033*** 0.116208*** 0.000907 (0.002364) (0.002562) (0.002662) (0.024788) (0.008493) insp_l1 0.005286*** 0.002437 0.002613 0.00399 0.008463 (0.001895) (0.001986) (0.002057) (0.019111) (0.006935) ave_lobby_flag ‐0.626483 (1.641941) ave_log_lobby 0.160736 (0.381256) ave_log_lobby_others 0.601511** (0.270574) r2_p 0.048 0.172 0.262 0.293 0.303 Chi2 71.008 147.187 86.82 153.402 86.442 119.906 N 3365 3363 1815 1815 1847 1529

Number of violations

(1) (2) (3) (4) (5) (6) (7) IMR3 ‐201.914015***‐44.865815** ‐58.607561***‐26.256018 70.215473*** ‐168.229488***3.635887 (57.985184) (18.126750) (22.145276) (18.205482) (19.043935) (46.147970) (34.830399) flag_controller_lobby 33.609882 11.833157 17.694165 14.598119 13.316769 173.993370** 147.861247 (44.725862) (30.428300) (27.666523) (28.429594) (16.905731) (84.816567) (97.524210) flag_controller_lobby_l1 ‐20.796827 ‐6.432723 ‐8.887746 ‐12.974945 ‐1.489399 ‐161.698255* ‐114.459589 (40.797589) (21.852039) (21.558924) (19.215630) (21.663604) (96.963574) (82.158717) log_lobby 12.695994 5.680191 6.182352 6.601464 ‐1.109752 27.358291 ‐3.364589 (8.814604) (4.433423) (4.135954) (4.382949) (2.265863) (50.519284) (29.500921) log_lobby_l1 ‐2.427746 ‐2.194421 ‐1.654963 ‐5.174439 1.896221 ‐11.94469 0.162681 (5.566028) (5.468230) (5.472781) (5.197954) (3.744217) (68.746124) (54.777277) log_lobby_others 0.047515 27.965813** 27.366172** 13.575318 38.529236** ‐13.252337 ‐9.589423 (9.385962) (11.787589) (11.603124) (21.626260) (19.126591) (122.983332) (129.021088) log_lobby_others_l1 ‐21.936755 ‐29.728179** ‐31.566209** ‐83.457154** ‐12.711272 ‐51.625918 ‐32.69509 (18.267707) (14.930180) (14.721514) (39.262828) (27.416640) (331.705564) (347.669318) Mineract 28.357266** 37.513766***39.958965*** 62.456855*** 16.748543 37.002494 17.339085 (12.281619) (12.643407) (12.658156) (21.522010) (18.918355) (149.982793) (149.553561) Insp 3.166388*** 3.301462*** 3.396764*** 2.822227*** 1.71025 2.577089 (0.573125) (0.581603) (0.610876) (0.678757) (2.372041) (3.028403) insp_l1 0.508024 0.618703 0.647515 0.243424 ‐4.044032* ‐2.663622 (0.472997) (0.529675) (0.546331) (0.597059) (2.078714) (1.939495) r2_a 0.123 0.618 0.622 0.632 0.726 0.421 0.674 F 5.389 29.515 43.741 61.25 71.016 26.217 100.389 N 1529 1529 1529 1529 1529 1529 1529

Total fines due

(1) (2) (3) (4) (5) (6) (7) IMR3 ‐3.821014*** ‐2.525022***‐2.471906*** ‐1.901456*** 1.521975*** ‐2.402268*** 1.066896** (0.532431) (0.341163) (0.364732) (0.325174) (0.366617) (0.391098) (0.458860) flag_controller_lobby 0.698096 0.518709 0.461731 0.235997 0.044055 1.472861 0.58547 (0.464126) (0.418485) (0.413652) (0.329391) (0.186128) (0.902665) (0.595387) flag_controller_lobby_l 1 0.259867 0.378083 0.378633 0.327931 0.225424 ‐1.295054 0.068529 (0.418830) (0.305156) (0.305999) (0.307717) (0.196789) (0.877708) (0.518531) log_lobby 0.102951 0.045009 0.046905 0.082682** 0.019839 0.931491* 0.071141 (0.091375) (0.050720) (0.048606) (0.040292) (0.035107) (0.508593) (0.421467) log_lobby_l1 ‐0.089857 ‐0.087895 ‐0.087487 ‐0.134907* 0.048716 ‐0.766997 0.05845 (0.085264) (0.092605) (0.093290) (0.075357) (0.045310) (0.713592) (0.497763) log_lobby_others 0.098794 0.329407*** 0.326563*** ‐0.061418 0.843126*** 0.091072 1.007097 (0.099123) (0.095931) (0.095667) (0.230911) (0.160948) (1.200469) (1.032087) log_lobby_others_l1 0.027766 ‐0.036799 ‐0.034691 ‐1.370083*** 0.531590* ‐1.622754 1.436072 (0.144087) (0.120925) (0.119687) (0.504896) (0.320186) (3.273834) (2.560516) Mineract 0.324664** 0.400206*** 0.390382*** 0.944372*** ‐0.131928 0.967306 ‐0.661659 (0.142808) (0.136093) (0.136859) (0.266192) (0.179656) (1.464945) (1.120244) Insp 0.026211*** 0.024178*** 0.024734*** 0.019844*** 0.028777 0.017098 (0.004147) (0.004319) (0.004201) (0.004064) (0.023012) (0.024827) insp_l1 0.004112 0.002152 0.001794 ‐0.004313 ‐0.046185*** ‐0.025029 (0.003255) (0.003629) (0.003788) (0.004539) (0.013972) (0.016485) r2_a 0.368 0.634 0.639 0.676 0.817 0.628 0.804 F 19.976 74.099 56.83 59.695 43.497 48.543 40.427 N 1529 1529 1529 1529 1529 1529 1529

Probability of fine reduction

(1) (2) (3) (4) (5) (6) flag_controller_lobby 0.104993** 0.048864 0.059739 ‐0.010673 0.055911 0.305657 (0.049088) (0.048942) (0.121502) (0.102093) (0.335354) (0.255810) flag_controller_lobby_l1 0.07373 0.085118 0.074179 0.040245 0.050822 ‐0.353706 (0.064354) (0.059525) (0.089959) (0.079477) (0.316870) (0.231784) log_lobby 0.004757 0.000035 0.004381 0.009725 ‐0.01698 0.200601 (0.010974) (0.009972) (0.015761) (0.014855) (0.058223) (0.127275) log_lobby_l1 ‐0.017732 ‐0.018304 ‐0.02197 ‐0.013433 0.007279 ‐0.14382 (0.012085) (0.011467) (0.015388) (0.014050) (0.057159) (0.184240) log_lobby_others 0.054674* 0.086609*** 0.100351*** 0.127698* 1.055325*** 0.251788 (0.028116) (0.026162) (0.035356) (0.074321) (0.278220) (0.343742) log_lobby_others_l1 ‐0.020426 ‐0.042855 ‐0.050909 ‐0.014403 0.828454 0.19781 (0.032970) (0.033562) (0.043989) (0.186211) (0.549348) (0.940742) Mineract 0.053525* 0.083126** 0.105573** 0.098826 0.024127 ‐0.014506 (0.030946) (0.034605) (0.046741) (0.098695) (0.277083) (0.417937) Insp 0.006408*** 0.004601** 0.004601*** 0.027317*** 0.006931 (0.001773) (0.001802) (0.001670) (0.006542) (0.007415) insp_l1 0.000218 0.000499 0.000664 ‐0.002077 ‐0.009778 (0.001488) (0.001672) (0.001499) (0.006387) (0.006483) ave_lobby_flag ‐0.485528 (0.525367) ave_log_lobby 0.101964 (0.095311) ave_log_lobby_others 0.324244** (0.164922) r2_p 0.04 0.175 0.21 0.241 0.221 Chi2 66.105 155.301 117.38 186.852 191.853 160.068 N 3365 3363 1815 1815 1847 1529

Granger Causality Tests

Inspections and Violations Y = insp (1) (2) Y = viol (1) (2) L.Insp 0.972373*** 0.943473*** L.viol 0.945930*** 0.843657*** (0.003969) (0.006091) (0.007231) (0.010924) L.Viol 0.008082*** L.insp 0.630415*** (0.001297) (0.051286) r2_a 0.946 0.947 r2_a 0.834 0.841 N 3396 3396 N 3398 3396 Violations and Fines Y = viol (1) (2) Y = log_due (1) (2) L.Viol 0.945930*** 0.898971*** L.log_due 0.915918*** ‐2.998595 (0.007231) (0.010887) (0.008792) (1.859777) L.log_due 4.884440*** L.viol 0.886676*** (0.849564) (0.023833) r2_a 0.834 0.836 r2_a 0.762 0.466 N 3398 3396 N 3396 3396 Violations and Lobbying Y = viol (1) (2) Y = log_lobby (1) (2) L.Viol 0.945930*** 0.944632*** L.log_lobby 1.017396*** 1.011254*** (0.007231) (0.007355) (0.007454) (0.007561) L.log_lobby 0.392645 L.viol 0.000609*** (0.406504) (0.000137) r2_a 0.834 0.834 r2_a 0.846 0.847 N 3398 3398 N 3398 3398 Standard errors in parentheses Significance levels: 10% * 5% ** 1% ***

Granger Causality Tests

Inspections and Fines Y = insp (1) (2) Y = log_due (1) (2) L.Insp 0.972373*** 0.960445*** L.log_due 0.915918*** 0.861296*** (0.003969) (0.004991) (0.008792) (0.010973) L.log_due 0.325823*** L.insp 0.005401*** (0.082959) (0.000660) r2_a 0.946 0.947 r2_a 0.762 0.766 N 3396 3396 N 3396 3396 Inspections and Lobbying Y = insp (1) (2) Y = log_lobby (1) (2) L.Insp 0.972373*** 0.971023*** L.log_lobby 1.017396*** 1.014872*** (0.003969) (0.004006) (0.007454) (0.007497) L.log_lobby 0.113614** L.insp 0.001682*** (0.047158) (0.000637) r2_a 0.946 0.947 r2_a 0.846 0.847 N 3396 3396 N 3398 3396 Fines and Lobbying Y = log_due (1) (2) Y = log_lobby (1) (2) L.log_due 0.915918*** 0.909120*** L.log_lobby 1.017396*** 1.002415*** (0.008792) (0.009020) (0.007454) (0.007549) L.log_lobby 0.020989*** L.log_due 0.093716*** (0.006388) (0.010658) r2_a 0.762 0.762 r2_a 0.846 0.85 N 3396 3396 N 3398 3396 Standard errors in parentheses Significance levels: 10% * 5% ** 1% ***