Saturday, August 20, 2005

The World of BLS Injury Rates: Where Every Employer is Better than the Average

The World of BLS Injury Rates:  Where Every Employer is Better than the Average

The BLS injury rate measures the risk of injury for an individual worker.  It compares that risk between industry sectors and facility sizes.  Data quality problems with unaudited employer reports aside, there’s some value to that number.  

The BLS number distinctly not appropriate for it’s most common use, which is inferring safety performance of a facility by comparing the facility’s incidence rate to the national “average” or some multiple of the national average.  Facilities get awards, recognition, exclusion from scheduled OSHA inspections for this.  Legislation is frequently introduced to reduce OSHA penalties.

The “average” or “arithmetic mean” is appropriate for data which follow a normal distribution.  Half the sample will have values above, and half below the mean, which will approximate the median or midpoint of the distribution.  Facility injury rates are distinctly skewed, so the arithmetic mean is distinctly inappropriate.

The distribution of incidence rates across facilities by industry sector and size is published annually by the Bureau of Labor Statistics.  The most recent data are for 2003, posted at  

Employers who manage more than three-fourths of American worksites report no injuries at all.  Because these data are displayed only in quartiles, we can’t estimate whether the no-injury cut point is 85th percentile, 90th percentile, 95th percentile or higher.  The national “average” [all private employement] is 5.0 workers injured and recorded per 100 employees.  Obviously the average (mean) is far above the median.  Since the median is also zero, the ratio of the average to the median is indeterminate.   For workplaces with less than 50 employees, the median is zero, with the no-injury cut point somewhere between 50th percentile and 75th percentile.  The next larger size ranges are 50-249, 250-999, and greater than 1000.  For each of these, the average is 128%, 145% and 157% of the median.  

For the manufacturing sector as a whole, where the national average is 8.1 workers injured and recorded per 100 employees, the median injury rate is zero, the 75th percentile facility carries an incidence rate well below the average.   The no-injury cut point is somewhere between the 50th and 75th percentile, and no ratio of average to median can be calculated for workplaces with less than 50 employees.  For the next larger size ranges, the average is 124%, 125% and 171% of the median.

We note that distributions for recorded Lost Workday case rates and recorded LWD cases with days away from work are even more skewed that the rates for total recordable cases.

The take home lesson is that being better than the national “average” is a pretty poor benchmark for a criterion for facility.

No comments: