WILLIAMS, Circuit Judge.

An administrative law judge disbelieved Anne Hill's testimony that she could not sit, stand, or walk for extended periods of time and denied her application for Disability Insurance Benefits and Supplemental Security Income. In this action arising under 42 U.S.C. § 405(g), Hill challenges this adverse credibility finding as well as the ALJ's assessment of her residual functional capacity. We agree with Hill that the ALJ's credibility analysis was flawed and remand the case to the agency for further proceedings.

I. BACKGROUND

A. Evidence of Hill's Disabilities

Hill, who is currently 56 years old, worked for more than 13 years at a steel factory, where her duties included lifting and carrying steel sheets that weighed up to 100 pounds. The manual labor took a toll on Hill's body, and, unable to keep working at the factory, she applied for disability benefits in July 2011. She alleged an onset date in June 2011 and listed eight impairments: a total hip replacement, a recommended total shoulder replacement, carpal tunnel, a ruptured disc, cervical fusion,1 knee pain, a broken left hand, and tendinitis.

Hill had cervical-fusion surgery in 1985, but there is no mention of neck pain in her available medical records before November 2010. In November 2010, she was diagnosed with neck strain, though she was cleared for work without restrictions. The next month, she saw a doctor for shoulder pain. Imaging studies revealed a likely hyperextension injury; a bone spur; tears of the tendons in her left shoulder, left bicep, and left hip joint; a possible "loose body" in her left shoulder; osteoarthritis in her left hip; and tendinopathy2 in her left shoulder. She was prescribed pain medication and her arm was placed in a sling. During a follow-up appointment with orthopedic surgeon Dr. Barry Liechty, Hill reported improvement but also said that she still experienced pain in her shoulder.

Hill met with Dr. Liechty again in May 2011, complaining of pain in her left hip, knee, and groin. Dr. Liechty noted severe osteoarthritis of the left hip and performed a total hip replacement two months later. A few weeks after that surgery, Dr. Liechty reported that Hill was limping and taking one or two Vicodin each week. He recommended that she not work at a job requiring pulling, pushing, or squatting. He also recommended that she not lift more than 10 pounds, but that recommendation was changed a few weeks later to restrict only "heavy lifting."

Nearly two months after her hip replacement, in August 2011, Hill met with state-agency doctor David Ringel, an osteopathic physician, who noted that Hill limped and had "quite a bit of stiffness." Dr. Ringel reported that Hill said she could lift 10 pounds, do most household chores, and stand 1 to 2 hours at a time and 4 hours total during an 8-hour workday. He described Hill as mildly obese and noted some limitations in her ability to move her shoulders, neck, lower back, and left hip.

The next month, in September 2011, another state-agency doctor, Dr. J. Sands, reviewed Hill's medical records and prepared a Physical Residual Functional Capacity Assessment. Dr. Sands noted Hill's history of shoulder and hip osteoarthritis. He opined that her "hip replacement would be expected to improve with further therapy and time," but said that she would be limited to occasional pulling, pressing, and pushing with her left extremities. He estimated that Hill could lift 20 pounds occasionally and 10 pounds frequently, and could sit, stand, or walk 6 hours each in an 8-hour workday. Dr. Sands further opined that Hill occasionally could balance, stoop, kneel, crouch, crawl, and climb ramps and stairs, but could not climb ladders, ropes, or scaffolds.

Two weeks later Hill complained of low back pain, and an imaging study revealed minimal degenerative disc disease, narrowing disc spaces, atherosclerotic3 vascular changes, and calcification of a portion of the pelvis. Dr. Sands and two additional state-agency physicians reviewed this new information but concluded that it did not change the earlier assessment of Hill's impairments.

The Social Security Administration initially denied Hill's application in October 2011, and did so again on reconsideration the next month.

In September 2012, Hill testified before the ALJ. She explained that she had to stop working at the factory because, after her hip-replacement surgery, she could no longer perform manual labor. Hill said that, despite taking aspirin and Naproxen, she could "hardly sleep at night" because of shoulder and neck pain, and she had trouble walking because of leg and back pain. But, Hill explained, it was difficult for her to determine the source of the pain because "[i]t's all connected somehow." She added that she no longer took narcotic pain relievers because her doctor was concerned that she could become addicted. (That concern may have been caused by Hill's admission of "heavy" drinking and a family history of alcoholism.) Hill then explained that she lacked health insurance and could not afford to go a doctor "over every little pain," nor could she afford the total shoulder replacement her doctor had recommended. She described her daily activities, which then included babysitting, caring for her pets and her roommate (who is an amputee), going to church, visiting with family members, and doing chores like loading the dishwasher, vacuuming, taking out the garbage, and doing laundry. But she added that she needs to take breaks while performing those chores and that she is unable to lift the child she babysits. She estimated that she could sit or stand for about 10 to 15 minutes without a break, lift between 10 and 15 pounds with her right arm but much less with her left, squat or kneel with her right leg but not with her left, and slowly crawl. Hill rated her hip pain at 3 out of 10, her shoulder pain at 8, and her leg pain at 5.

A longtime friend, Kim Stamate, testified that she helps walk Hill's dogs and carry Hill's groceries, and that she had not seen Hill walk without a limp in the past year.

A vocational expert testified that Hill could not perform her past work if limited to the extent described by the ALJ: able to lift up to 20 pounds occasionally and 10 pounds frequently; able to sit, stand, or walk for 6 hours each in an 8-hour workday; occasionally able to balance, stoop, kneel, crouch, crawl, and climb ramps and stairs, but never able to climb ladders, ropes, or scaffolds; and limited to occasional pulling, pushing, and reaching with the left extremities. But, the VE continued, Hill still could work at jobs classified as light4 and unskilled, such as a dealer account investigator,5 a furniture rental consultant, and a counter clerk.6 In the national economy, the VE said, there were 22,000 jobs as a dealer account investigator, 40,000 jobs as a furniture rental consultant, and 50,000 jobs as a counter clerk. Those jobs, the VE asserted, would be available to a person who was unable to crouch or to push, reach, or pull with her left extremities. Finally, the VE testified that someone with Hill's limitations, if also unable to stand or walk more than 2 hours in an 8-hour workday or lift more than 10 pounds occasionally, still could perform sedentary jobs, see SSR 83-10, 1983 WL 31251, at *5; 20 C.F.R. §§ 404.1567(a), 416.967(a), like a call-out operator7 (45,000 jobs in the national economy), a semiconductor bonder8 (30,000 jobs in the national economy), and a registration clerk9 (27,000 jobs in the national economy).

B. The Agency's Decision

The ALJ concluded that Hill is not disabled. The ALJ applied the 5-step analysis for assessing disability, see 20 C.F.R. §§ 404.1520(a)(4), 416.920(a)(4), and first determined at Step 1 that Hill had not engaged in substantial gainful activity since her alleged onset in June 2011.

At Step 2 the ALJ identified Hill's severe impairments as degenerative joint disease with total replacement of the left hip and osteoarthritis of the left shoulder. The ALJ also concluded, however, that Hill's alleged back and neck pain do not cause "more than minimal functional limitations" and "do not further reduce her capacity beyond the limitations imposed by her left hip and left shoulder impairments." The ALJ reasoned that, because Hill had worked at a job involving manual labor "for many years" after her neck surgery and had a full range of motion in her neck in December 2010, her neck pain likely had resolved before her alleged onset date in June 2011. As for Hill's back, the ALJ noted that the imaging study done in October 2011 had "revealed only minimal degenerative changes," which, according to the ALJ, "would not be expected to cause significant pain without any nerve root impingement or spinal stenosis." What's more, the ALJ added, Hill had "not had any epidural injections, physical therapy, or other treatment for low back pain," did not have an abnormal gait, and had "not been evaluated by a specialist for low back pain."

At Step 3 the ALJ concluded that none of Hill's impairments, individually or in combination, satisfy a listing for presumptive disability.

At Step 4 the ALJ rejected Hill's account of the severity of her disabling limitations. The ALJ started with boilerplate language reciting that Hill's "statements concerning the intensity, persistence and limiting effects of these symptoms are not entirely credible" even though her "medically determinable impairments could reasonably be expected to cause the alleged symptoms." The ALJ then asserted that Hill's hip pain had improved significantly since her surgery: Hill had decreased her use of pain medication; she did not use a walker; her doctor had cleared her for work with minimal limitations; she had not complained in her most recent visits with doctors of hip, leg, or back pain; and her doctor had reported that Hill's "gait was not abnormal." Turning next to Hill's alleged back pain, the ALJ repeated that an imaging study had shown "only minimal degenerative changes in the lumbar spine" and that "[w]ithout any nerve root compression, loss of disc space height, or spinal stenosis, the extent of low back pain and inability to sit, stand, and walk alleged by the claimant is not credible, particularly in light of the absence of complaints of low back pain to treating physicians." And, concerning Hill's neck pain, the ALJ again asserted that the lack of any complaints of pain to doctors after November 2010 "suggests that the neck pain resolved quickly."

Concerning Hill's shoulder, however, the ALJ acknowledged that Hill's testimony was supported by evidence of osteoarthritis and tendinopathy in the medical records. But the ALJ concluded that, although Hill's use of her left arm was limited, she still could use her right arm to lift and carry 20 pounds occasionally and 10 pounds frequently. Hill's ability to complete some housework, the ALJ insisted, confirmed that her use of her right arm was not limited. And, the ALJ added, Hill's use of only aspirin and Naproxen for pain "suggests that the pain is not as limiting" as she testified.

The ALJ gave great weight to the medical opinions of Hill's treating orthopedist and the state-agency physicians who suggested that she could perform a limited range of light work. But the ALJ discounted the testimony of Hill's friend, Stamate, reasoning that she was not a "trained medical observer" and had contradicted the observations of Hill's doctor. The ALJ noted also that Hill "wanted to work but could not find someone to hire her."

Finally, at Step 5 the ALJ concluded that Hill could not perform her past relevant work but could perform light work, including as a dealer account investigator, furniture rental consultant, or counter clerk. Finding those jobs existed in significant numbers, the ALJ found Hill not disabled.

The Appeals Council denied review, making the ALJ's decision the final decision of the Commissioner. See Pepper v. Colvin, 712 F.3d 351, 361 (7th Cir.2013). The district judge upheld that decision.

II. ANALYSIS

In this court Hill challenges the adverse credibility finding, arguing that the ALJ improperly discredited her testimony that back and neck pain limit her ability to stand, sit, and walk for extended periods of time. We agree. As discussed below, the ALJ's analysis was flawed in several respects.

First, the ALJ reasoned that Hill's credibility was undermined because she had stopped taking narcotic pain relievers and had complained of back pain to doctors only intermittently. Yet the ALJ ignored explanations for the conservative treatment: Hill testified that her doctor was worried about the addictiveness of narcotic pain relievers and that her back and neck pain may have been related to her shoulder and hip pain, which she did complain about to doctors on multiple occasions. See SSR 96-7P, 1996 WL 374186, at *7 (ALJs must consider "any explanations that the individual may provide, or other information in the case record, that may explain infrequent or irregular medical visits or failure to seek medical treatment"); Beardsley v. Colvin, 758 F.3d 834, 840 (7th Cir.2014) (remanding to agency where ALJ made no attempt to determine reason for conservative treatment). The ALJ acknowledged Hill's testimony that she could not afford to visit doctors frequently, but nonetheless found Hill's testimony about back pain not credible because Hill "did not even seek a referral to a back specialist." How Hill would have paid for a specialist, the ALJ did not say. See Craft v. Astrue, 539 F.3d 668, 679 (7th Cir.2008) (ALJ should have considered claimant's "inability to pay for regular treatment and medicine").

Second, the ALJ reasoned that Hill was exaggerating her back pain because she had never been diagnosed with nerve root compression, loss of disc space height, or spinal stenosis. The ALJ's conclusion is not supported by any medical evidence in the record; it amounts to the ALJ improperly "playing doctor." See Engstrand v. Colvin, 788 F.3d 655, 660-61 (7th Cir.2015); Goins v. Colvin, 764 F.3d 677, 680 (7th Cir.2014); Pate-Fires v. Astrue, 564 F.3d 935, 946-47 (8th Cir.2009).

Third, the ALJ reasoned that Hill was stretching the truth about her neck pain because she still wanted to work, and because she performed manual labor for many years after her neck surgery. This logic is backward: a "claimant with a good work record is entitled to substantial credibility when claiming an inability to work because of a disability." Rivera v. Schweiker, 717 F.2d 719, 725 (2d Cir.1983); see Singletary v. Sec'y of Health, Educ. & Welfare, 623 F.2d 217, 219 (2d Cir.1980) (claimant's history of performing demanding work over long hours "justifies the inference that when he stopped working he did so for the reasons he testified to"); Allen v. Califano, 613 F.2d 139, 147 (6th Cir.1980). And, as we have explained, a claimant's desire to work is not inconsistent with her inability to work because of a disability. See Voigt v. Colvin, 781 F.3d 871, 876 (7th Cir.2015) (claimant's desire to work, but inability to find work, is "consistent with his wanting to lead a normal life yet being unable to land a job because he's disabled from gainful employment"); Jones v. Shalala, 21 F.3d 191, 192 (7th Cir.1994) (explaining that claimant might be earning a decent wage despite being permanently disabled).

The Commissioner counters that consideration of a claimant's work history is "proper when the claimant has had essentially the same condition for decades, and remained able to work." That statement is true enough, see Orlando v. Heckler, 776 F.2d 209, 213-14 (7th Cir.1985); Genier v. Astrue, 606 F.3d 46, 49 (2d Cir. 2010); Johnson v. Apfel, 240 F.3d 1145, 1148-49 (8th Cir.2001), but Hill has never asserted that the severity of her impairments have remained the same for decades. Hill contends instead that she gradually experienced worsening health problems that eventually became disabling in June 2011, shortly before she applied for benefits. Indeed, the ALJ found that Hill suffers from degenerative joint disease (osteoarthritis), which often grows more severe with the passage of time. See Roddy v. Astrue, 705 F.3d 631, 637 (7th Cir.2013) (ALJ's reasoning flawed because it ignored the fact that "degenerative" conditions get worse over time).

At oral argument, counsel for the Commissioner attempted to liken Hill's activities of daily living, particularly babysitting, with working full time. But Hill's home life was not a reason given by the ALJ for denying benefits and so not a proper basis for us to uphold the ALJ's decision. See Sec. & Exch. Comm'n v. Chenery Corp., 318 U.S. 80, 87, 63 S.Ct. 454, 87 L.Ed. 626 (1943); Hanson v. Colvin, 760 F.3d 759, 762 (7th Cir.2014). (The ALJ addressed Hill's home activities only to the extent that they showed that she had use of her right arm.) Counsel's reasoning is unsound, in any event, because we have repeatedly warned against equating the activities of daily living with those of a full-time job. See Bjornson v. Astrue, 671 F.3d 640, 647 (7th Cir.2012); Spiva v. Astrue, 628 F.3d 346, 352 (7th Cir.2010).10

It might be argued that these mistakes in evaluating Hill's credibility were harmless. As the Commissioner points out, no doctor has opined that Hill has more limitations than the ALJ incorporated into her assessment of Hill's residual functional capacity. But Hill testified that she is more limited, and her testimony cannot be disregarded simply because it is not corroborated by objective medical evidence. See Hall v. Colvin, 778 F.3d 688, 691 (7th Cir.2015); Pierce v. Colvin, 739 F.3d 1046, 1049-50 (7th Cir.2014); Robbins v. Soc. Sec. Admin., 466 F.3d 880, 883 (9th Cir.2006). We are not confident that the ALJ would have reached the same conclusion about Hill's credibility had she not inappropriately "played doctor," ignored possible explanations for Hill's conservative treatment, and conflated a desire to work with the ability to do so. So the ALJ's errors are not harmless. See McKinzey v. Astrue, 641 F.3d 884, 892 (7th Cir.2011) (explaining that error is harmless only if court is "convinced" that ALJ would reach same result on remand); Spiva, 628 F.3d at 353 (error is not harmless simply because ALJ might have reached the same result absent the error).

III. CONCLUSION

We REVERSE the district court's judgment and REMAND this case to the Commissioner for further proceedings.

POSNER, Circuit Judge, concurring.

I join Judge Williams's majority opinion without reservations. I write separately only to focus attention on what seems to me a persistent, serious, and often ignored deficiency in opinions by the administrative law judges of the Social Security Administration denying social security disability benefits (or, what is similar, supplemental security income). The deficiency concerns testimony by vocational experts employed by the Administration concerning the number and types of jobs that an applicant deemed not to be totally disabled could perform, and the evaluation of that testimony by administrative law judges. This deficiency has recently been attracting critical attention. See, e.g., Browning v. Colvin, 766 F.3d 702, 708-09 (7th Cir.2014); Herrmann v. Colvin, 772 F.3d 1110, 1113-14 (7th Cir. 2014); Jon C. Dubin, "Overcoming Gridlock: Campbell After a Quarter-Century and Bureaucratically Rational Gap-Filling in Mass Justice Adjudication in the Social Security Administration's Disability Programs," 62 Administrative Law Review 937, 964-71 (2010); Peter J. Lemoine, "Crisis of Confidence: The Inadequacies of Vocational Evidence Presented at Social Security Disability Hearings (Part II)," Social Security Forum, Sept. 2012, p. 1, www.lemoinelawfirm.com/wp-content/uploads/2012/seminar_materials.pdf (visited November 25, 2015). Expert opinions must not be "conjured out of whole cloth." Brault v. Social Security Administration, 683 F.3d 443, 449-50 (2d Cir. 2012) (per curiam), citing Donahue v. Barnhart, 279 F.3d 441, 446 (7th Cir. 2002).

One problem addressed in the cases and academic commentary that I've cited is that the only reliable statistics on number of jobs are census data of broad job categories, rather than data on the number of jobs in the narrower categories of jobs that the applicant for benefits could actually perform. Typically, it appears, the vocational expert simply divides the census estimate of the number of jobs in the broad category that includes the narrow category of jobs that the applicant can perform, by the total number of narrow categories in the broad category. The assumption is thus that every narrow category has the same number of jobs as every other narrow category within the broad category—a preposterous assumption.

Usually the administrative law judge, toward the end of his or her (in this case her) opinion denying benefits (the government as a matter of policy never appeals the grant of benefits), states that the limits to the applicant's ability to work are not so severe as to render him or her incapable of full-time gainful employment (in which event, with the exception noted below, the applicant would be deemed totally disabled and entitled therefore to benefits). The administrative law judge then tells the vocational expert what the applicant can and cannot do and asks the expert to opine on the nature and number of jobs the applicant can perform that exist in the economy. If there is a nontrivial number of such jobs, the applicant is denied benefits. (Were the non-triviality condition not satisfied, it would be plain that the applicant was incapable of full-time gainful employment and so would be entitled to benefits.) The administrative law judge's list of what the applicant in this case can and can't do is: "can lift up to 20 pounds occasionally, lift and carry up to 10 pounds frequently, stand or walk for approximately 6 hours per 8 hour day, and sit for approximately 6 hours per 8 hour workday, with normal breaks ..., occasionally climb ramps and stairs, balance, stoop, kneel, and crawl, but never climb ladders, ropes, or scaffolds, crouch, push or pull with the left upper and left lower extremities, or reach in all directions (including overhead) with the left upper extremity." The applicant is 56 years old, worked for more than 13 years at a steel factory where her duties included lifting and carrying steel sheets that weighed up to 100 pounds, and is acknowledged to be unable to do such work anymore because the manual labor that she performed for so many years took a toll on her body.

The administrative law judge found that the applicant has "severe impairments" consisting of "degenerative joint disease with total replacement of the left hip and osteoarthritis of the left shoulder." (She has many subsidiary impairments as well, mainly on her left side like the severe impairments.) It seems ridiculous to think she can stand or walk for six hours in an eight-hour workday or "balance" (whatever that means), or do full-time work that even "occasionally" involves stooping, kneeling, crouching, crawling, or climbing ramps (depending on the angle—try climbing a ramp that is inclined by 45 degrees). Given the impairments of her left extremities (arm and leg) it is doubtful that she can pull, push, or reach with them; at least this issue central to her application required reasoned analysis, not an unexplained conclusion (though I acknowledge the severe time pressures under which the Social Security Administration's administrative law judges labor). Nor was there any reason for the administrative law judge to say that the applicant can't hold down a job that involves climbing ropes, as that is beyond obvious (and what kind of jobs in today's economy involve climbing ropes?).

The limits of the applicant's ability to use her "left extremities" bothered the vocational expert. He testified that the Dictionary of Occupational Titles does not describe jobs that can be performed with only one fully functioning limb, and he therefore had to rely on his "own experience" for his conclusion that the applicant can perform light and unskilled, or sedentary and unskilled work. But he failed to describe the experience that formed his opinion, despite our criticism in Herrmann v. Colvin, supra, 772 F.3d at 1113, of the vocational expert who failed to explain "how impressions from unspecified past experience and `knowledge' could enable him to determine numbers of particular jobs" that persons with specified physical or mental impairments could perform. The question of how to incorporate a limb restriction into a description of work capability in the Dictionary of Occupational Titles has been a recurring issue, e.g., Carey v. Apfel, 230 F.3d 131, 145-47 (5th Cir.2000), and remains unresolved.

As examples of jobs that the applicant could perform despite the "left extremities" problem, the vocational expert instanced dealer account investigator, furniture rental consultant, and counter clerk, and opined that in the national economy there are 22,000 jobs of the first type, 40,000 of the second, and 50,000 of the third. But having said that, he quickly hedged his statement that such jobs are within the capacity of a person unable to crouch or to push, reach, or pull with her left extremities, by acknowledging that "the difference between the dominant hand and non-dominant hand is not actually specified in the DOT." He said he was basing his testimony instead on his "experience on how those jobs are actually performed," though, as we said, he never explained what that experience was.

And finally he said that someone with the applicant's than two hours in an 8-hour workday or lift more than 10 pounds occasionally could still perform sedentary jobs like a call-out operator (45,000 jobs in the national economy), a semiconductor bonder (30,000), and a registration clerk (27,000).

His testimony collapses when one considers the nature of the listed jobs. He had said on the basis of the administrative law judge's description of the applicant's limitations that the applicant could do only unskilled "light" or "sedentary" work. Here are how the six jobs he said the applicant can perform are described in the Dictionary of Occupational Titles. A dealer account investigator "visits dealers to verify purchases financed by [a] bank against physical inventory of merchandise." That doesn't sound like unskilled work. Nor does "furniture rental consultant." For his definition of "counter clerk" the vocational expert cited the paragraph in the Dictionary of Occupational Titles that defines processing film: the counter clerk "receives film for processing, loads film into equipment that automatically processes film for subsequent photo printing, and collects payment from customers of photofinishing establishment." (That might actually be a job that the applicant in this case could do—the only one in the list.) A call-out operator "compiles credit information, such as status of credit accounts, personal references, and bank accounts to fulfill subscribers' requests, using telephone." Sedentary yes, unskilled no. A semiconductor bonder tends an "automatic bonding machine that bonds gold or aluminum wire to integrated circuit dies to connect circuitry to package leads." Again, sedentary but not unskilled—and it could be dangerous for someone with a nonfunctioning left arm. Last is registration clerk, and here the vocational expert seems to have lit on the wrong paragraph of the Dictionary of Occupational Titles. The paragraph he cites refers to an election clerk—someone who performs administrative tasks "during elections." Obviously that is occasional rather than full-time employment, because elections are not held continuously.

In short, the vocational expert's testimony was worthless—and this apart from the apparent arbitrariness of his numerology. It is time the Social Security Disability Office cleaned up its act.