The problem presented in data sets is one of correlation between two variables -- the independent and dependent nature of the variables. To illustrate the problem, we examine Marker (DOT 369.687-026). Labor assigns Marker to Inspectors, Testers, Sorters, Samplers, and Weighers (SOC 51-9061). Labor assigns 782 DOT codes to this group, 173 of which are unskilled.
The 2020 Occupational Outlook Handbook states that 557,900 jobs function as Inspectors. The 2020 Occupational Employment and Wage Statistics estimates 549,200 jobs as Inspectors. The 2021 Occupational Requirements Survey states that 18.1% of the jobs are unskilled. The ORS estimates that 30% of inspectors engage in medium work.
If there is no correlation between skill level and exertion -- the data points are independent of one another -- the number of unskilled medium jobs is 5.43% of the total number of jobs, less than 30,000 jobs. If the correlation between skill and exertion is 1:1, all unskilled jobs require medium exertion, then the number of medium unskilled jobs is 18.1% of the total number of jobs, less than 100,000 jobs. The correlation is neither zero nor 1:1. The typical vocational expert lacks the expertise to state the correlation or interdependence. There are likely unskilled sedentary, light, and heavy jobs.
Adding the variable of standing (which includes walking) adds another layer of complexity. The ORS states that Inspectors engage in sedentary work in 10.4% of jobs and light work in 48.6% of jobs. That leaves roughly 12% of jobs as indeterminate. It is also worth noting that Labor has redefined light work as including lifting and carrying up to 25 pounds occasionally.
It is clear that 75th and 90th percentiles capture all of the sedentary jobs, sitting 75% to 90% of the day. That range also captures jobs that require more than sedentary exertion because the ORS reports 10.4% of jobs require sedentary exertion. Some of the jobs require light and might require higher levels of exertion. Sitting and sedentary exertion have a correlation that is embedded in the definition of sedentary work. The question turns on the other 90% of jobs.
The 75th and 90th percentiles for standing describe Inspectors as standing 85% or more of the workday. The 50th percentile states that Inspectors stand 75% of the workday, six hours of an eight-hour day. The question is whether sitting 25% of the workday at the median correlates to or is dependent on the exertion otherwise expended: light; medium; or heavy. The Commissioner defines light and medium work the same, standing six hours and may sit intermittently during the remaining time in SSR 83-10. The ruling permits the presence of light work with long sitting but does not permit the presence of medium work sitting most of the day. While this blog has argued that SSR 83-10 is wrong and entitled to no deference or respect on the standing/walking issue, unskilled medium sit-down work probably does not exist.
The inference that flows suggests that some indeterminate range of jobs above the 25th and below the 75th percentiles permit standing/walking less than or equal to 75% of the day or six hours. The 75th percentile and above require more than six hours of standing. The 25th percentile (unstated by estimated from its mirror image of sitting at the 75th percentile) requires prolonged sitting.
Do we know the correlation or dependence ratios? Not for medium and light work, we just don't without Labor giving us raw data that would take a statistician to understand. But, our clients do not have the burden of production or proof at step five, the existence of other work. So here is the suggested methodology:
1. Ask the VE to identify the applicable and replicable methodology.
2. Ask the VE to state whether the methodology is well-accepted.
3. Ask the VE to assume that the OOH, OEWS, and ORS are accurate. Assuming that evidence is true, ask for a statement of how the testimony given is consistent with the data published by the Department of Labor.
4. Ask the VE to explain how the data points presented in the hypothetical question correlate -- are these data points dependent or interdependent.
We actually want a failure of the VE to be able to answer question 4. That permits us to argue that the ALJ should treat them as wholly independent. That permits if not requires as cascaded application of the job numbers by each factor, one after another.
This is complicated and most will not feel comfortable the first dozen attempts. Like anything difficult, it will take practice and perfection. We can, we must, climb the mountain.
___________________________
Suggested Citation:
Lawrence Rohlfing, Beware of Independent versus Dependent Variables in Data, California Social Security Attorney (November 29, 2021) https://californiasocialsecurityattorney.blogspot.com
