X hits on this document

60 views

0 shares

0 downloads

0 comments

11 / 25

Metrics: (1) the response of a plant’s compliance status to lagged EPA/state enforcement and monitoring activities directed at that facility, (2) the response of a plant’s pollution emissions to lagged EPA/state enforcement and monitoring activities directed at that facility. Metrics should be explored on a sector-by-sector basis.

Peer-Reviewed Foundation: Gray and Deily (1996), Gray and Shadbegian (2005), Gray and Shadbegian (2007), and Deily and Gray (2007).

Potential Data Requirements: (1) compliance status (a discrete 0/1 indicator variable) or specific pollutant emissions (a continuous variable) for plant i in time period t,6 (2) a year indicator for time t, (3) a season indicator for time t if the data are monthly or quarterly, (4) a state indicator for plant i, (5) inspections at plant i over the past year, (6) inspections at plant i 1-2 years ago, (7) inspections at plant i 2-3 years ago, (8) enforcement actions at plant i over the past year, (9) enforcement actions at plant i 1-2 years ago, and (10) enforcement actions at plant i 2-3 years ago.

Potential Statistical Methodologies for the Continuous Emissions Metric: (1) linear regression with plant-specific fixed effects, (2) linear regression with plant- specific random effects, and (3) linear regression with plant-specific conditional random effects. The continuous nature of the emissions metric suggests linear regressions are appropriate. All methods are discussed in Section 2 and the technical appendix of the Task 1 white paper. All are easily implemented (pre- programmed) with modern statistical software. All models include state-specific indicator variables and year-specific indicator variables.

7

Potential Statistical Methodologies for the Compliance Status Metric: (1) logit regression with plant-specific fixed effects, (2) logit regression with plant- specific random effects, and (3) logit regression with plant-specific conditional random effects. The discrete nature of the 0/1 compliance status metric suggests non-linear logit regressions are more appropriate than linear regression models. All methods are discussed in Section 2 and the technical appendix of the Task 1 white paper. All are easily implemented (pre-programmed) with modern statistical software. All models include state-specific indicator variables and year-specific indicator variables.

Key Simplification and Justification 1: Detailed variables representing plant and community characteristics assembled from non-EPA datasets are omitted. State indicator variables, year indicator variables, and panel data statistical

6 In principle, compliance status may refer to any desired compliance indicator, including agency determined Significant Non-compliance status or High-Priority Violation status. In the Gray and co-author papers that serve as the foundation of the later specific deterrence benchmarking analyses, compliance status is EPA-determined as reported in the Compliance Data System and the more recent Integrated Data for Enforcement Analysis Database.

7 Technically, models with plant-specific fixed effects will not also include state-specific indicators, since these variables are redundant.

10

Document info
Document views60
Page views60
Page last viewedMon Dec 05 13:05:50 UTC 2016
Pages25
Paragraphs327
Words9837

Comments