Opinion
19-cv-1019-jdp
07-31-2020
JOHN WESTENDORF, Plaintiff, v. ANDREW M. SAUL, Commissioner, Social Security Administration, Defendant.
OPINION and ORDER
Plaintiff John Westendorf seeks judicial review of a final decision of defendant Andrew M. Saul, Commissioner of Social Security, finding him not disabled under the Social Security Act. The administrative law judge (ALJ) found that Westendorf had the residual functional capacity to perform sedentary work with some additional physical restrictions. The ALJ adopted the testimony of the vocational expert that a person with those limitations could work as a document preparer, address clerk, and call-out operator, and that all these jobs are available in significant numbers in the national economy.
On appeal, Westendorf asserts that the job-number estimates provided by the vocational expert (VE) were not based on reliable methods. The VE drew her estimates from a commercially available software product, but she could not explain how the software generated the estimates. The VE submitted documentation from the software provider about the product's methodology, but that documentation was out of date and it didn't adequately explain how the software worked. The court concludes that the VE's estimates aren't supported by substantial evidence, so it will remand this case to the agency for further proceedings. The oral argument scheduled for August 4, 2020, is cancelled.
BACKGROUND
For a disability claimant to qualify for benefits under the Social Security Act, the ALJ must find that the claimant's impairments preclude him from performing full-time work. The ALJ first assesses whether the claimant's limitations preclude him from performing jobs he held in the past. 20 C.F.R. §§ 404.1520(f) and 416.920(f). If, as in this case, the ALJ finds that the claimant cannot perform his past work, the commissioner has the burden to establish that there are a significant number of jobs in the national economy that someone with the claimant's abilities and limitations would be able to perform. 20 C.F.R. §§ 404.1560(c)(2) and 416.916(c)(2).
In this case, the commissioner relied on the testimony of a VE, who testified that a person with Westendorf's limitations could work as a document preparer, address clerk, or call-out operator, and that respectively, these jobs have 46,500, 5,695, and 5,240 positions nationwide. R. 26. When asked about how she derived those numbers, the VE testified that she used "SkillTRAN, Job Browser Pro," R. 74, which is a commercial software program that provides job numbers for specific occupations by DOT code. See Dahl v. Saul, No. 18-C-676, 2019 WL 4239829, at *4 (E.D. Wis. Sept. 6, 2019). She testified that Job Browser Pro was "the standard system [used by] VEs all over the United States and judges, attorneys, and anyone else that works with people . . . looking for work." R. 74.
Record cites are to the administrative transcript, located at Dkt. 9. --------
When questioned by Westendorf's attorney, the VE testified that Job Browser Pro's numbers are derived from data from the Bureau of Labor Statistics, R. 75, and that she uses the software's default industry codes when generating job numbers, R. 76. But she was unable to explain how Job Browser Pro actually derives DOT-code-specific job estimates from that data. She said that there was "a six-page document that shows the methodology on how they came about the specifics of those statistics," which she promised to provide after the hearing. Id. Westendorf's attorney made a conditional objection to the VE's job estimates as unreliable pending an opportunity to review that document. R. 77.
The VE provided the methodology document. R. 393-98. Westendorf objected in a letter to the ALJ, raising three concerns about the reliability of Job Browser Pro's methodology. First, he noted that the software automatically "'suggests' various industries for each DOT [occupation]" and that the ALJ had "used these default suggested industries," but that there was no information "to establish the reliability of the industries selected by SkillTran. . . . The VE did not testify whether she has reviewed the 'suggested' industry codes to determine whether these codes are reliable." R. 407.
Second, Westendorf asserted that Job Browser Pro relies in part on the "equal-distribution method," which has been repeatedly criticized as unreliable by the Court of Appeals for the Seventh Circuit. See Chavez v. Berryhill, 895 F.3d 962, 969 (7th Cir. 2018) (collecting cases). Under this method, DOT code-specific job estimates are generated using data from the Department of Labor's compilation of Occupational Employment Statistics. But that publication doesn't use the DOT job-grouping system; instead, it relies on a separate classification system, the Standard Occupational Classification (SOC). The SOC job estimates reflect broader categories of jobs and may encompass any number of DOT job titles. Under the equal-distribution method, DOT-specific job estimates are calculated by dividing the total number of jobs in a SOC category by the number of specific DOT job titles that fall within that category. Id. at 965. This method almost certainly results in inaccurate estimates because it rests on the flawed assumption that all job titles within a category exist in equal numbers.
Third, Westendorf objected to the VE's assertion that Job Browser Pro "is used by VEs, judges, and attorneys 'all over the United States'" as conclusory and unsupported. R. 408. He noted that she had given "no information pertaining to how widely used this source is, whether it is widely accepted, or the basis upon which she made her statements regarding its use 'all over the United States.'" Id.
In his decision, the ALJ overruled Westendorf's objection. He noted that "the vocational expert provided a full explanation of her use of Skilltran/Job Browser Pro" and then repeated the substance of the VE's testimony without addressing the specific concerns Westendorf had raised:
[The vocational expert] said [that Job Browser Pro] provided a basis to make an estimate of numbers of jobs and it is based on the Department of Labor's Bureau of Labor Statistics information for its source. The vocational expert fully explained the methodology and said that Skilltran/Job Browser Pro is commonly used in the vocational consultant community as a source. The vocational expert further said that Skilltran/Job Browser Pro is a source that is available for public use.R. 26. The ALJ made passing reference to Job Browser Pro's methodology, asserting that "the vocational expert did in fact identify sources/methodology sufficient for the undersigned to find her testimony well supported and credible." Id. But he did not specifically discuss the methodology document provided by the VE or otherwise explain why he was satisfied that the methodology used to generate the job estimates was well-supported.
Having overruled Westendorf's objection, the ALJ relied on the VE's job estimates to find that Westendorf could perform available work and was not disabled. R. 27.
ANALYSIS
On appeal, Westendorf reasserts that the method used by the VE to produce her job estimates was flawed. To withstand appeal, an ALJ's decision must be based on substantial evidence. Summers v. Berryhill, 864 F.3d 523, 526 (7th Cir. 2017). If the decision relies on a VE's job estimates, it is based on substantial evidence only if those job estimates are the product of a reliable method. Chavez, 895 F.3d at 968. This isn't an exacting standard, as it's impossible to avoid some amount of uncertainty in developing such estimates. Id. But "any method that the agency uses to estimate job numbers must be supported with evidence sufficient to provide some modicum of confidence in its reliability." Id. at 969.
Westendorf contends that job estimates obtained through Job Browser Pro are generated using the disfavored equal-distribution method and are therefore presumptively unreliable. Although the Seventh Circuit has not expressly enjoined use of the equal-distribution method, the implication of its ruling in Chavez is that a VE needs to justify the reliability of estimates produced using that method when their accuracy is called into question. See id. at 970 (reversing ALJ decision based on job estimates calculated using the equal-distribution where "nothing in the administrative record allow[ed the court] to conclude with any reliability that the estimates reasonably approximate the number of suitable jobs that exist" for the claimant). Westendorf says that the VE failed to provide such assurances here.
In response, the commissioner denies that Job Browser Pro relies heavily on the equal-distribution method. Quoting the six-page methodology document provided by the VE, the commissioner asserts that it is only at a "much smaller, subset level"—a level that does not occur in all instances—that Job Browser Pro "initially assume[s] equal distribution of employment." R. 395. In light of that, the commissioner notes that it is "not at all clear that equal distribution occurred in this case, as the vocational expert's statement that she used the 'default codes' does nothing to indicate whether the data at issue was part of the much smaller subset where equal distribution occurs." Dkt. 15, at 5 n.3. The commissioner nonetheless assumes throughout his brief that the equal-distribution method was used in some way because, in his view, the estimates would be reliable regardless.
This argument puts the commissioner in the awkward position of defending the reliability of a method that he's not even sure was employed in this case. It also illustrates the more fundamental problem: the Job Browser Pro methodology document is insufficient to show the VE's methods were reliable, for two fundamental reasons.
First, the methodology document is out of date. The VE testified that she was using the current version of the program, and that she had updated it just before the hearing in 2018. But the methodology document is dated December 26, 2008, apparently shortly after the public release of the software. The document explains that the software will be revised and improved, so it's not clear whether the 2008 methodology document accurately describes the software that the VE used in 2018.
Second, the methodology document is opaque and does not fully describe how the software works. The software is intended to address the problem of the equal-distribution method. R. 393. The software relies heavily on Occupational Employment Statistics from the U.S. Bureau of Labor Statistics, which is a perfectly acceptable source of data. But the methodology document does not fully explain how it uses the OES data to produce estimates of the number of positions available within jobs defined in the DOT. The document itself acknowledges that, as of 2008, the software was not yet ready to withstand a challenge under Daubert standards, R. 397, which federal courts use to evaluate the admissibility of scientific and technical evidence. Although the VE's evidence in a disability hearing need not pass muster under full-blown Daubert scrutiny, see Chavez, 895 F.3d at 968, the documentation shows that Job Browser Pro was a new "disruptive" approach that had not been thoroughly tested or widely accepted. Job Browser Pro was simply a black box to the VE. And even with the methodology document, it was likewise a black box to the ALJ.
The ALJ did not address Westendorf's substantive objections to the software, nor did the ALJ actually evaluate the method as described in the documentation. The ALJ simply parroted the VE's assertions that Job Browser Pro relies on government data, that it is commonly used by vocational consultants, and that it is available for public use. None of that is shows that the methodology is reliable. After all, the exact same things can be said of the equal-distribution method. Nothing in the record supports the conclusion that the VE's use of Job Browser Pro in this case produced reasonably reliable job estimates. By accepting the VE's conclusory assurances of reliability without explanation or analysis, the ALJ failed to "hold the VE to account for the reliability of [her] job-number estimates." Id. at 970.
The commissioner suggests that this court conclusively determined that Job Browser Pro employs reliable methods in Kenealy v. Saul, No. 19-cv-40-jdp, 2019 WL 6463840 (W.D. Wis. Dec. 2, 2019). That is incorrect. In Kenealy, the VE gave a more fulsome account of how she generated her estimates: she compared estimates derived from Job Browser Pro with estimates derived through the equal-distribution method and then arrived at ballpark estimates based on her experience and expertise. The plaintiff made a general reliability objection, but she didn't take specific exception to the VE's use of Job Browser Pro, so that wasn't the focus of the court's analysis. Likewise, the Seventh Circuit has yet to opine on the reliability of Job Browser Pro's software. District courts in this circuit have affirmed decisions that rely on Job Browser Pro estimates where VEs were able to meaningfully describe the processes they employed. See, e.g., Chad S. v. Saul, No. 1:18CV298, 2020 WL 773274, at *5-6 (N.D. Ind. Feb. 18, 2020); Khuzaie v. Comm'r of Soc. Sec., No. 1:14-CV-00199-SLC, 2016 WL 1253537, at 15 (N.D. Ind. Mar. 30, 2016). But they have remanded in cases where VEs provided conclusory explanations of their processes and the ALJ failed to substantively address objections to Job Browser Pro's reliability. See, e.g., Smith v. Berryhill, No. 16 C 5451, 2018 WL 4679584, at *7-8 (N.D. Ill. Sept. 28, 2018); Russell v. Berryhill, No. 16-cv-1251, 2017 WL 3704354, at *4-6 (S.D. Ill. Aug. 28, 2017).
The commissioner makes two additional arguments in an effort to rehabilitate the VE's estimates. First, he asserts that the VE "explained that, in addition to utilizing SkillTRAN Job Browser Pro, she reduced the job number estimates 'based upon [her] vocational opinion, from [her] experience, training, and observation of jobs, and how jobs are actually performed in the national economy.'" Dkt. 15, at 4 (quoting R. 74, 78). The commissioner correctly notes that a VE can support her estimates by "drawing on knowledge of labor market conditions and occupational trends, gleaned from reviewing relevant data sources or from placing workers in jobs." Chavez, 895 F.3d at 970. But the parts of the hearing transcript that the commissioner cites are drawn from the VE's testimony explaining why she reduced the job estimate specifically associated with the job of "marker"—a light-work job that the ALJ didn't ultimately rely on in finding Westendorf not disabled. At no point did the VE indicate that she made reductions in the document preparer, address clerk, or call-out operator job estimates, or indicate that that she was adjusting those estimates based on her training and experience. So that argument misses the mark.
Second, the commissioner suggests that remand isn't warranted because Westendorf hasn't shown that the VE's estimates are inaccurate or that the use of some other methodology would have yielded a different outcome. But it was not Westendorf's burden to make such a showing. See id. ("By accepting the VE's estimates at step five because they were 'not contradicted,' the ALJ effectively and impermissibly shifted the burden to Chavez."). Westendorf raised reasonable concerns about the reliability of the VE's methods, and the ALJ failed to substantively address those concerns at the hearing or in his decision.
Because the ALJ did not ensure that the VE's job estimates were the product of a reliable method, his decision was not based on substantial evidence. On remand, the ALJ will need to inquire into the VE's methods, whether those methods are reliable, and whether they were accurately applied to the case at hand before relying on the VE's conclusions. Job Browser Pro might be a useful tool, but the VE needs to be able to explain how Job Browser Pro makes its job number estimates, how she used the software to generate her own estimates, and why she believes those estimates are reliable.
ORDER
IT IS ORDERED that the decision of Andrew M. Saul, Commissioner of Social Security, denying plaintiff John Westendorf's applications for disability insurance benefits and supplemental security income is REVERSED AND REMANDED under sentence four of 42 U.S.C. § 405(g) for further proceedings consistent with this opinion. The August 4, 2020 oral argument is CANCELED as unnecessary.
Entered July 31, 2020.
BY THE COURT:
/s/_________
JAMES D. PETERSON
District Judge