Pennsylvania certified safety committee program

Region was based their injury rates. The state was divided into three areas of the impact of financial incentives for firms to adopt safety according to the Federal District Court system: eastern, committee programs. There are several methodological central, and Western Pennsylvania. We included the region advantages in this study: we have a relatively large set of variable because we thought that awareness of the program observations; we can look at changes in injury rates might vary in different parts of the State.

Program over time; we have relatively valid measures of whether compliance measures for audited firms were derived from firms complied with program requirements and which audits, such as presence of a safety committee, composition requirements they were; we applied propensity score of the committee, safety and injury records, minimum matching to overcome the self-selection problem.

Perhaps training of committee members, and frequency of the most importantly, we have two different data sets, which committee meetings. This method is more informative Data sources than a model that simply examines whether a firm ever joined.

Program participation was modeled as a function of Several data sources were used for this study. Employ- firm size, the number of workplaces, the prior injury rate, ment data came from the Pennsylvania Unemployment industry, and region. We focused first on firms that joined Insurance system UI spanning the years —, the program in and later since we have injury rate data including all the firms operating in Pennsylvania. The injury beginning only in Since a to The BWC also provided random audit results firm generally behaves consistently in its decision to join the of CSC participants since , including compliance program, its participation status in 1 year may be correlated information on 22 separate areas of possible deficiencies.

We estimated a logistic In addition, we used a file with inspection data — model using the generalized estimating equations method to from the federal Occupational Safety and Health account for the correlation between years as we are interested Administration OSHA Integrated Management Informa- in the population-level effects; and the correlation structure tion System IMIS.

Both the OSHA inspection data and the within firms was tested using quasi-likelihood information CSC data include information on the presence of a union, criteria [Pan, ; Hardin and Hilbe, ].

We reported the allowing us to examine the effect of union status on program estimates based on a model with the best correlation structure participation and impact on injury rates. These datasets were linked together at the firm level, To explore other possible explanations, we did several based on unique firm numeric identifiers or matching using a additional analyses.

If independent variables in the model. The purpose was to a firm had multiple establishments, the information was examine whether the change in the injury rate matters in the aggregated to the firm level. Key variable definitions 2 Analysis 3: We tested a model to predict whether a firm ever participated in the program between and , by Throughout the analysis, the injury rate was defined as aggregating the information across years.

Some to one non-participant that had the closest propensity to that insurers may have advertised the program less heavily or of the participant one-to-one nearest neighbor matching. The top 10 insurers in the state were identified and propensity, one of them was randomly selected as the control. These insurers were We used one-to-one nearest neighbor matching because it identified based on average total premiums in and achieved good balance between the two groups regarding [Pennsylvania Department of Labor and Industry, ].

Non-participating analysis on firms which had ever participated in the CSC firms that were matched in earlier years were excluded in the program and tried to identify the factors that predicted matching process for later years. Only one participant was program dropout. Since the actual date of dropping out is not outside the region of common support, meaning that its clear in the data, we only examined whether a firm was in the propensity was higher than the maximum or less than program by the end of The analysis was limited to the the minimum propensity of the non-participants, and was firm-years after participation using the data from to excluded from the final analysis.

After matching, the firm We used a simple logistic model and the variables in the level characteristics were not significantly different between model were aggregated at the firm level across multiple years. The implementation on injury rates. We first compared, for each regression model included program participation, firm size, audit criterion, the unadjusted change in rates for compliant the number of workplaces, and industry.

We included and non-compliant firms. We then selected measures only firms that had at least 1 year of injury data prior to that were significantly different between two groups and participation in the program in order to have a more valid examined their possible role in explaining changes in injury comparison of injuries before and after participation.

This analysis answered the Analysis 8: However, because participation in the question of whether the variation in program implementation program is voluntary, firms that participated in the program affects work injuries.

As an alternative, we RESULTS employed propensity score matching to compare the injury rate changes for the participants to the changes for a group of Descriptive Statistics non-participating firms whose probability of participation was the same. Thus the control group becomes a set of very The final analytical sample for statistical analysis similar firms which, for an unknown reason, decided not to consisted of , firms and 1,, firm years over participate [Dehejia and Wahba, ; Caliendo and the period between and Figure 1 shows the time Kopeinig, ].

Overall participation in the Since firms joined the program at different time points, it program increased over time, growing from less than 0. On average, participating the participants and non-participants experienced the firms in our sample participated in the program 6.

First, a logistic regression was run for 11 possible years. Again, the independent participants see Table I. Participating firms were much ing firm based on the logistic regression. We found that most of the same factors that affected the probability of participation also affected the probability of dropping out Analysis 6. In addition, among audited firms, those not complying with any requirements were over 16 times as likely to drop out after the audit as firms with at least some compliance.

Participation inthe certifiedsafety committee program,by year. Table III shows the regression analysis for the sample that includes all participants and non-participants Analysis versus Most non-participating injuries per workers.

The mean injury rate of participants workers higher than the rates for non-participants. The was 2. Firms with over employees were almost times participants, a decline of 0. Firms decline of 1. For both with injury rate ratios greater than 0. In the construction sector, the advantage for controlling for other factors. Compared to the firms in the non-participant firms was even larger. Audited Firms Additional analyses showed that the level of the injury rate change in the prior 2 years had no effect on participation Among the audited firms out of that could be Analysis 2.

Examining whether a firm ever participated in matched to the injury data, 79 had a safety committee that met the program during through found that the all the requirements at the time of audit compliant group , probability of participation was higher among larger firms 18 had no committee at all non-compliant group , and and firms with higher injury rate ratios Analysis 3. Firms firms had some violations with an average of 2. Firms insured by certain top 10 WC insurers had group.

On average, the compliant group had a higher injury a lower rate of participation than those insured by other rate 4. The numbers reflect average injury rates between and The geographic distribution of firms is not reported in this table in order to save space.

In other words, the compliant group the non-compliant group had increased. These were the had a decrease in injury rate, while the rates of the other requirements that 1 an agenda be prepared for each groups increased. In a bivariate analysis, we detection, accident investigation, and other safety and health compared Table V the average pre-program injury rate issues specific to the firm. For further regression analysis, we and the average post-program injury rate for firms with dropped the requirement for records on meeting attendance and without each of the audit deficiencies Analysis 9.

The next three coefficients indicate that firms which were later found to have these deficiencies tended to have had lower pre-CSC injury rates could not play a major role in explaining the variation in than compliant firms or firms with other deficiencies. Again, injury rate changes. Along with the two requirements above, Firms with the citation for absence of training for committee we also included a variable indicating that a firm had no members demonstrated a 1.

The without having to do anything different appears not to have controls were matched with participants based on propensity score as described in been the dominant factor. Many RAND studies are published in peer-reviewed scholarly journals, as chapters in commercial books, or as documents published by other organizations.

Our mission to help improve policy and decisionmaking through research and analysis is enabled through our core values of quality and objectivity and our unwavering commitment to the highest level of integrity and ethical behavior. To help ensure our research and analysis are rigorous, objective, and nonpartisan, we subject our research publications to a robust and exacting quality-assurance process; avoid both the appearance and reality of financial and other conflicts of interest through staff training, project screening, and a policy of mandatory disclosure; and pursue transparency in our research engagements through our commitment to the open publication of our research findings and recommendations, disclosure of the source of funding of published research, and policies to ensure intellectual independence.

For more information, visit www. The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. Find a Job. Find Services. Individual Services. Disability Services.

Employer Services. Certified Safety Committee Information Providing Pennsylvania workers with a safe working environment is a major priority. Committees must be in existence and operating for six 6 FULL, consecutive calendar months prior to signing, dating and submission of the Application.

A quorum of committee members must meet at least monthly. The Application must be received by the Workers' Compensation Certification Unit, in Harrisburg, between 90 and 30 calendar days in advance of the next workers' compensation policy renewal date.