Quick Answer: Research conducted by Myvesta (founded by Steve Rhode) found that 49.3% of people in debt crisis screened positive for depression symptoms on the CES-D — about half of the 136 people who walked in asking for help with a math problem. For years this page compared that to an unsourced 9.5% and called it a five-fold risk. That comparison was built wrong — it set a screening result against a diagnosis rate — and I show below how I rebuilt it from a single peer-reviewed source. Rebuilt properly it is a range, not a number — roughly two to five times, on a survey of 136 people. I show the whole working below, including where the old figure came from and my own wrong turn along the way. I believe the finding still helps explain why “grind it out” debt advice fails so many people who ask for help — though a screen is not a diagnosis, and nothing here can prove which way the causation runs.
What this research changed about how I work: I built a full guide to money anxiety around this finding — including the half of it that debt advice never addresses, where the numbers are fine and the fear is not.
Corrected 3 August 2026. For twenty-five years this page ran a comparison that does not hold up, and I want you to see that before you read anything else. It set my 2001 screening result against a general-population diagnosis rate — two different measurements — and wrongly called the gap five-fold. I have rebuilt it from a single peer-reviewed source. The honest answer is a range, roughly two to five times, on a survey of 136 people. The full correction is below, including the two attempts I got wrong first.
Debt and Depression: The Myvesta Foundation Research
This page documents original research conducted by Myvesta (formerly Debt Counselors of America), the nonprofit credit counseling organization founded by Steve Rhode in 1994. This study remains one of the few clinical examinations of depression rates among people in debt crisis.
Why This Research Matters: Most debt advice assumes people can simply “buckle down” and follow a plan. This research is why I think that assumption fails for a large share of people in debt crisis. I believe depressive symptoms make it harder to act; I believe the inaction that follows feeds the debt cycle.
Study Overview
Methodology
The study used the CES-D (Center for Epidemiologic Studies Depression Scale), a validated clinical screening instrument developed by the National Institute of Mental Health. The CES-D is widely used in research and clinical settings to identify individuals at risk for clinical depression.
- Participants: 136 Myvesta clients, surveyed August 2001. Unsecured debts from $1,000 to more than $100,000; ages 21 to 77; incomes from $6,000 to $165,000. Real completed CES-D responses from people in debt crisis — not a model, an estimate or a projection
- Published: Debt, Stress and Depression Survey, Myvesta.org, released November 2001. The original 14-page report is here — read it yourself rather than taking my word for any of this. I have left that historical document exactly as it was published in 2001, which means it still contains the 9.5% comparison I correct below
- Instrument: CES-D depression scale (20-item questionnaire)
- Comparison: results were originally set against a general-population figure that, as I explain in the correction below, was not measuring the same thing
- Scoring: Standard CES-D cutoff scores for depression symptoms
Key Findings
Overall Depression Rates
| Group | Screened positive (CES-D) | In the severe range |
|---|---|---|
| Myvesta debt crisis clients (2001) | 49.3% | 39.7% |
This page used to carry a second row here — a 9.5% “general population” figure — and a stat card calling the difference a five-fold risk. Both were wrong, for a reason worth understanding, and I deal with it in the correction below rather than quietly deleting it.
Critical Finding: Nearly 40% of people seeking debt help scored in the severe range of the screen — not mild sadness, but the symptom load a clinician would want to look at. A screen is not a diagnosis, and I am not diagnosing anyone; what it means is that four in ten people arriving for help with a budget were carrying enough symptoms that someone should have asked a follow-up question — and for almost every one of them, nobody ever did.
Gender Differences
The 136 people were 81 women and 55 men; 47 of the women and 20 of the men screened positive. That gap does survive a conventional statistical test, which genuinely surprised me when I finally ran it — I had assumed a sample this small could not support the claim the page was making. Two cautions belong with it. It is a difference in screening results, so the crisis-symptom problem I describe further down applies to it as much as to the headline figure. And gender was one of several breakdowns the 2001 report published, none of them set out in advance as a hypothesis — so treat it as something worth looking at, not as something established.
Income Level Analysis
Every income band in the survey contained people who screened positive. That is genuinely all this data can tell you about income, and I want to be careful even with that. If the spread between bands is noise — and it is — then pointing at the top bracket and calling it meaningful is the same error running the other way. The survey cannot tell you whether income protected anyone. It can only tell you that no bracket came back empty.
What I will not do any more is rank those bands against each other, because this page used to, and it should not have. The survey split 136 people across seven income brackets, which leaves some cells holding as few as nine people. At that size a bracket moves twenty points if two people answer differently. The screened-positive rates across the seven bands run from 36% to 78%, and that spread is not a finding about income — it is what happens when you cut a small survey into small pieces.
A specific correction: for years this page wrongly claimed a similarity between income groups, and anchored it to a “$100,000” bracket. I am deliberately not restating that sentence in a readable form anywhere on this page — not even struck through — because a crossed-out line still reads as an assertion to a search snippet or a screen reader. Two things were wrong with it. There is no $100,000 bracket in the study at all. The highest band is “over $90,000,” and it holds eleven people. The $100,000 figure came from a different sentence in the 2001 release, describing who our clients were — “many of our clients have incomes exceeding $100,000 a year” — which I turned into a comparison the data never made. It is the same mistake as the 9.5%: a number lifted out of one context and presented as a measurement.
The Debt-Depression Cycle
This is what I believe I watched, and not something my survey proved. I believe depressive symptoms make it harder to act; I believe the debt then worsens; and I believe the symptoms deepen from there. I think that is why “grind it out” advice fails so many people. A one-time survey cannot establish which end of that came first, and I am not going to pretend mine did.
— Steve Rhode, Founder of Myvesta
What I watched looked like a self-reinforcing cycle. I want to be careful with the wording, because this is the pattern I believe I saw across thousands of clients — not something a single snapshot survey can establish as cause and effect:
- I saw financial stress and depressive symptoms show up together — worry, shame, hopelessness — and which of the two arrives first is exactly what this survey cannot tell you
- Those symptoms appear to make deciding harder — difficulty making decisions, difficulty following through on a plan
- Not acting tends to worsen the finances — missed payments, accumulating interest, collection calls
- Worse finances seem to deepen the symptoms — and, in what I watched, the loop closes. This is my read of a pattern, not a proven mechanism
Implications for Debt Advice
These are my conclusions, not the survey’s. The 2001 study measured depression symptoms and nothing else. It did not test a single one of the approaches below — those come from thirty years of watching what helped people and what did not.
What Works
- Acknowledging the emotional reality of debt
- Breaking tasks into small, manageable steps
- Addressing mental health alongside financial strategy
- Considering faster resolution options (bankruptcy) when appropriate
- Removing shame from the conversation
What Fails
- “Just follow the budget” advice
- Long-term repayment plans requiring sustained motivation
- Shame-based approaches (“you got yourself into this”)
- Ignoring the emotional component
- Assuming willpower alone is sufficient
Why Steve Rhode Founded Myvesta
Steve Rhode filed personal bankruptcy in 1990 after his real estate business collapsed in an economic downturn. This lived experience—combined with the recognition that debt is primarily an emotional problem—led him to found Debt Counselors of America in 1994 (later renamed Myvesta).
At its peak, Myvesta employed 70 people including:
- Staff psychologists
- Lawyers
- Mediators
- CPAs and tax experts
- Employment specialists
Myvesta was the first organization to launch an inpatient program for compulsive spending, recognizing that debt problems often have deeper psychological roots.
Key Takeaways
- Of 136 people in debt crisis surveyed in 2001, about half screened positive for depression symptoms and four in ten scored in the severe range — a large elevation over the general population, best stated as a range of roughly two to five times rather than the single five-fold figure this page carried for years
- Four in ten scored in the severe range of the screen. The survey never measured whether daily functioning was impaired, so I do not claim that it was
- Every income band in the survey contained people who screened positive — but the per-band numbers are far too small to rank against one another
- The debt-depression cycle may explain why willpower-based advice fails — that is my reading of what I watched, not something a one-time snapshot can prove
- In my experience, effective debt help has to address the emotional reality and not just the financial math — that is my conclusion from practice, not a finding of this survey
How old is this study, and does it still hold?
I want to deal with the obvious objection myself, before someone else raises it. This research was conducted in 2001. That is a long time ago. So the fair question is not “is this a striking number” — it is “is this still true, and how would you know?”
What this study can and cannot tell you
What it can: that among the 136 people who contacted one nonprofit for debt help in 2001 and were screened with the CES-D, roughly half screened positive for depression symptoms and about 40% scored in the severe range. Every income band tested contained people who screened positive — though the bands are far too small to compare against one another.
What it cannot:
- It cannot tell you which caused which. This is a snapshot, not a longitudinal study. It cannot separate “debt made people depressed” from “depression made people worse with money” — and in my experience the honest answer is that both run at once, each feeding the other.
- It is a self-selected sample. These were people who had already decided they needed help and picked up a phone. That is not the same as a random sample of everyone in debt. My own expectation — and this is my inference, not a finding of the study — is that people at the point of asking for help are in worse shape than people quietly managing.
- A CES-D score is a screen, not a diagnosis. The CES-D flags people at risk for clinical depression. It does not diagnose anyone, and neither do I. The same caution applies to the Debt Stress Test on this site.
- It is one organization, in one country, in one year. Treat the precise decimals as directional. It is the size and consistency of the gap that matters, not whether the true figure is 49.3% or 47%.
A correction I owe you: the comparison itself was wrong
For years this page set 49.3% against 9.5% and called the difference a five-fold risk. I do not stand behind that comparison any more. Since the entire point of the section above is that I will tell you what my own research does and does not prove, the correction belongs here in plain sight rather than quietly disappearing from the page one night.
The two numbers were never measuring the same thing. My 49.3% is a screening rate — the share of people whose CES-D score cleared the threshold that flags someone as worth a closer look. It was being compared against a figure describing how many people are diagnosed with depressive illness. A screen and a diagnosis are not the same event, and a screen is deliberately built to over-catch, because the cost of missing a sick person is worse than the cost of a second conversation.
How much it over-catches is measurable. The largest review of this instrument — Vilagut and colleagues (2016), pooling 28 studies and 10,617 people in PLOS One — found that at the standard cutoff the CES-D has a sensitivity of 0.87 and a specificity of 0.70. In plain language: it catches about 87 of every 100 people who are genuinely depressed, and about 30 of every 100 people who are not depressed will screen positive anyway. The authors work the arithmetic themselves. Screen 10,000 people in a population where 10% genuinely are depressed, and you correctly catch 870 of them while also flagging 2,700 who are fine. Just under 36% of everyone screened comes back positive before you have gone looking in any unusual population at all.
Which is how you can tell the old number was impossible
You do not need to know where the 9.5% came from to see the problem. At the standard cutoff the false-positive floor alone is around 30%, so a general population screened this way does not land near 9.5% — the arithmetic will not produce it. I can now tell you exactly where it came from, because I went back to the original report. On page one of the November 2001 release, right under our own result, it says: “In comparison, studies have shown that 9.5 percent of the general population is clinically depressed.” No citation. No study named. The word doing the damage is “clinically.” We set our screening result against somebody else’s diagnosis rate, offered no source for it, and the comparison has been repeated ever since — by me, on this page, for years. That error is mine and it is twenty-five years old.
So what is the honest comparison? This is where I have to show you my own working, because my first attempt at fixing this was also wrong.
My instinct was to compare my 49.3% against that ~36% screening rate, which gives an elevation of about 1.4 times — and to tell you the real answer was “closer to double than five times.” Several people I asked to tear this apart caught the error, and they were right. Comparing two raw screening rates is itself misleading, because both numbers carry the same ~30% of false positives as dead weight. That shared ballast mechanically squashes any ratio toward 1. Vary the general population’s underlying rate across any plausible value and that raw-screen comparison lands between roughly 1.2 and 1.6 — not because the difference is small, but because the measurement is noisy on both sides.
One thing to be clear about before I go further. The 49.3% is a measurement. 136 people in debt crisis sat down and answered a twenty-item questionnaire and we counted the results: 67 of them scored as depressed, 54 of those in the severe range. Nothing in the argument below touches that count.
But 136 is a small survey, and I should have been saying so all along. At that size the margin of error around 49.3% is roughly plus or minus eight points — the true figure for the population my clients were drawn from could reasonably be anywhere from about 41% to about 58%. That is not a flaw in the survey; it is just what 136 people can tell you. It does mean every number below inherits that width, and the gender breakdown, which splits those 136 into two smaller groups still, is wider again.
The comparison that actually means something is between the underlying rates, with the false positives backed out of both sides. And there is a tidy way to do that without mixing sources, which is what caused this mess in the first place: take every number from the same review, so the accuracy figures and the comparison figure rest on the same reference standard.
That review reports a median prevalence of major depression of 8.8% across the general-population studies it pooled, with an interquartile range of 3.8% to 12.6% — all of it verified against structured diagnostic interviews, the same standard used to establish the accuracy figures above. Back the false positives out of my 49.3% using those two accuracy figures and it implies roughly 34% of my clients were genuinely depressed. Here is that step in full, so you can check it rather than trust it: true prevalence = (observed − (1 − specificity)) ÷ (sensitivity − (1 − specificity)), which is (0.493 − 0.30) ÷ (0.87 − 0.30) = 0.34. Against that 8.8% median that would be about four times. But run the same arithmetic across my own margin of error — 41% at the low end, 58% at the high end — and the implied elevation moves between roughly two and five and a half times. From sampling alone. Before anything else is accounted for.
So was the old number right after all?
Not quite, and the reason matters more than the number. It got there by the wrong route, and I would rather say that out loud than bury it. The original comparison was built wrong: it set a screening result against a diagnosis rate, which is not like for like and is not something I will defend. What I did not expect, when I set out to fix it, was that doing the arithmetic properly would land near where the page started. Around four times, against a five-fold claim. The old number turned out to be arithmetically reachable — and reachable is not the same as correct. The reasoning that produced it was wrong either way — and as you will see in a moment, even that four is an upper bound on the central estimate, not my best guess.
So what changed is narrower than a retraction and wider than a footnote: the method was wrong, and the magnitude is smaller than the old headline but still large. That is a less satisfying story than either “I was right all along” or “I made it up,” and it is what the numbers actually support.
Now the part I have to be straight about, because it cuts against the number I just gave you, and I would rather be the one to say it.
Four times is the top of that central estimate, not my best guess. Two things push it down, and both push the same way.
The first is the one that matters most, and it is specific to who these people were. My clients were not people with debt. They were people in crisis. Look at what the CES-D actually asks: whether your sleep was restless, whether your appetite was poor, whether everything you did felt like an effort, whether you could not get going. Now picture someone three months behind with the phone ringing all day. They will answer yes to those questions honestly and truthfully, and some of them will be describing their circumstances rather than a depressive illness. The accuracy figures I used to back out false positives were measured in ordinary populations who are not living that. In a group like mine, the test almost certainly threw more false positives than the 30% I subtracted — which means my 34% estimate is too high, and so is the four-fold that follows from it.
The second is smaller but runs the same direction. The CES-D asks about the past week. The diagnostic interviews behind that 8.8% comparison generally work to a longer window and screen out distress that is situational. Comparing the two tilts the ratio upward again.
So what do I actually believe?
That the elevation is real, that it is large, and that it is a range, not a number — somewhere in the region of two to five times, and I cannot honestly narrow it further. Those two numbers are doing different jobs, so let me be plain about it. The span is what sampling error on 136 people does to the estimate — the raw arithmetic runs from about two at the low end to about five and a half at the very top. The central estimate inside that span is the four I just called an upper bound. And the two effects I described push the whole span downward, not just its middle — which is why I will not defend the top of it either, and why I write two to five rather than carrying the decimal. Publishing a figure to one decimal place here would be false precision inside a correction about false precision. Narrowing it further would take a structured clinical interview nobody conducted in 2001, and I cannot invent that now.
Which means the old “more than five times” sat at the very top of the range and was quoted as though it were a fact. It was not fabricated — you can see where every piece of it came from, and now so can I — but it was the most flattering reading of a small survey, hardened into a headline. I am replacing it with the range, because the range is what the data supports.
What survives all of this without needing any arithmetic at all is the thing that actually mattered. About half the people who came to me for help with a budget screened positive for depression, and four in ten scored in the severe range. Whatever share of that was illness and whatever share was crisis, every one of those people was in genuine distress, nobody in the industry was checking, and I could not have told you which was which either — which is exactly why I hired people who could.
What survives the correction
The multiplier was wrong. The finding is not. Roughly half of the people walking through my door screened positive — in a group nobody had ever thought to screen, for a problem the whole industry insisted was arithmetic. I am deliberately not restating a multiple here, because comparing my raw screening rate against a general-population screening rate is the very move I spent the section above explaining is misleading. Whatever the true elevation is, it was large enough that I hired psychologists.
Vilagut and colleagues are blunt about the limit that applies to all of this. That review’s conclusion, word for word: “the accuracy of the CES-D is acceptable for its use as a first-stage screener to target respondents with depressive symptoms for more in-depth clinical assessment, but its use as a stand-alone measure for diagnostic purposes is not recommended.” That caution applies to my 2001 study exactly as it applies to the Debt Stress Test on this site. Neither one diagnoses anybody.
What the independent research has found since
Here is the part that actually answers the “is it still true” question, and it is the reason I have kept this page up rather than quietly retiring it. I did not have to defend this finding alone. In the twenty-five years since, a substantial independent literature has grown up around the same relationship — different countries, different instruments, different decades, researchers with no connection to me and nothing invested in my conclusion.
I keep the actual papers in a working reference library rather than relying on summaries of them. These are the ones that bear directly on this study, listed so you can go and read them yourself:
- Emily Ryu & Lu Fan (2022) — Financial Stress and Depression in Adults: A Systematic Review. The broad modern survey of what is established about the relationship.
- Shu Hui Ng, Azmawati Mohammed Nawi & Rozmi Ismail (2020) — Relationship Between Debt and Depression, Anxiety, Stress, or Suicide Ideation. A second systematic review, reaching into the most severe end of the range.
- Qing Hu, Ross Levine, Chen Lin & Mingzhu Tai (2019) — Mentally Spent: Credit Conditions and Mental Health. Examines credit conditions against measured mental-health outcomes, including antidepressant use — which sidesteps the self-report problem my 2001 study had.
- Kim (2021) — Financial Debt and Mental Health of Young Adults.
- Cude et al. (2014) — College Students and Financial Distress: Exploring Debt, Financial Satisfaction and Financial Anxiety.
- Smigielski et al. (2008) — Decision-Making and Risk Aversion among Depressive Adults. Speaks directly to the mechanism this page proposes: what depression does to the capacity to decide and follow through.
- MIDUS Study, University of Wisconsin (2020) — How Shame Intensifies Financial Hardship.
- Christopher P. Guzelian, Michael Ashley Stein & Hagop S. Akiskal (2015) — Credit Scores, Lending, and Psychosocial Disability.
- Liu et al. (2026) — Persistent Financial Adversity and Cognitive Aging: A Life Course Investigation, Innovation in Aging, following the 1946 British birth cohort. The longest lens anyone has yet put on this, and I wrote about it separately in financial stress and cognitive decline.
An honest note about how I am using these. I am listing them as further reading, not claiming that any one of them reproduces my 49.3% figure — none of them set out to, and I am not going to reduce somebody else’s careful work to a number that flatters mine. What the literature supports is the relationship: that financial hardship and depressive symptoms travel together, strongly and repeatedly, across populations and decades. My 2001 study is one early, small, self-selected data point inside that larger picture. It is worth reading because of what it caught early and because of what it changed about how I work — not because it is the last word.
Why I have not quietly retired this page
Because of what nobody was doing at the time. In 2001 I was running a credit counseling organization, and the entire industry — including the parts of it I competed with — was built on the premise that debt is an arithmetic problem and the feelings are a side effect. We screened our clients for depression because I could not make sense of what I was watching otherwise: people whose numbers were improving who felt no better, and people whose numbers were terrible who slept fine the night they finally had a plan.
Roughly half of the people walking through the door for help with a math problem were carrying something that would have surfaced the moment anyone thought to ask, and not one of them had been asked. That is the finding. It is why I hired psychologists, why Myvesta ran the first inpatient program for compulsive spending, and why I still will not tell someone to just follow the budget. Twenty-five years later I do not think the industry has fully absorbed it.
Where to take this next: if you recognized yourself anywhere in the above, the practical guide built on this research is Money Anxiety: which kind do you have? — including the half of the problem this 2001 study never looked at, where the numbers are objectively fine and the fear is there anyway. If debt has pushed you somewhere darker than worry, Debt and Mental Health has the crisis lines and the deeper material.
Citing This Research
When citing this research:
Citation: Myvesta.org, Inc. (2001). Debt, Stress and Depression Survey. Released November 2001. n=136 clients surveyed August 2001, screened with the CES-D. Results: 49.3% screened positive for depression symptoms (9.56% moderate, 39.71% severe); 50.74% not depressed. Elevation over the general population is best stated as a range of roughly two to five times, never a single multiplier. Full report (PDF).
Please cite it honestly: a screening rate in a small, self-selected sample of people already in debt crisis — not a diagnostic prevalence, and with a margin of error of roughly ±8 points. Do not repeat the “9.5% general population” comparison that appeared in the original release; it is unsourced and compares a screen to a diagnosis. See the correction on this page.
Source: GetOutOfDebt.org/research — Steve Rhode, founder of Myvesta (1994-2006)
Related Resources
- About Steve Rhode — Background on the researcher
- Bankruptcy Guide — When a fresh start makes sense
- Find Your Path Quiz — Personalized debt relief recommendations
