A preferred stock is an odd hybrid. It pays a fixed dividend the way a bond pays interest, and the big credit firms rate it the way they rate bonds. Yet it trades on a stock exchange and sits in ordinary brokerage accounts next to common shares. For income investors the appeal is plain: yields of 6% to 8%, often from financially sound companies. The trouble is that nobody has offered those investors a good way to judge one.
Moody’s, Standard & Poor’s and Fitch rate preferred securities, but their ratings answer an institutional question: how likely is the company to default, and how much would a lender lose if it did. They were never meant to say whether a particular preferred belongs in an individual’s income portfolio.
The big preferred-stock indexes bundle hundreds of issues into baskets. That helps someone choosing a fund. It does not help someone weighing two specific securities on a watchlist.
What research exists for individual investors mostly takes the form of checklists: a list of tests, each passed or failed, with the score a count of the passes. Checklists have real value, and they have brought discipline to many portfolios.
YieldDesk starts from a different premise. Two preferreds with the same score can be weak in entirely different places. One might be backed by a strong company but face redemption in a year. Another might have years of runway but a thin earnings cushion behind its dividend. A checklist gives both the same number. An investor deciding whether to hold, add or replace needs to know which weakness is which.
The Income Quality Score, or IQS, is our answer. It is a 0-to-10 grade computed as a weighted average of five measures, each itself scored from 0 to 10, with partial credit where partial credit is earned. A preferred that scores 8 lost its two points somewhere specific, and the product shows where, every time the score appears.
The IQS is in beta.The method is published and in use, but we are still tuning the weights and thresholds, and a security’s score can change for that reason alone. When it does, the movement says something about our tuning, not about the security. Treat the score as one input among several, not a settled verdict.
What the Income Quality Score measures
The IQS judges each preferred on five measures. Each is scored from 0 to 10 using the security’s own data, and the overall score is their weighted average, with coverage counting most. The five measures and their weights are summarized below, with the detail in the next section.
Two preferreds with the same score can be weak in entirely different places, and the difference matters, because what an investor should do next depends on where the weakness is.The YieldDesk thesis
The five measures, in detail
Coverage
The first and heaviest measure asks the simplest question: how comfortably does the company earn what it owes? A preferred is only as safe as the earnings standing in front of it. A pristine rating and generous terms count for little if profits barely clear the payment, because that is exactly the situation in which an income investor gets hurt.
What counts as “comfortably” depends on how the company makes its money, and this is the single largest change in version 3. For an industrial company or a utility, the test is pre-tax earnings against fixed charges: five times or better earns full credit, three times scores 8, two times 6, one times 3, and below that zero. For a bank, an insurer, or a mortgage company that funds itself with borrowed money, that ratio is meaningless: paying interest is the business itself, and even the soundest names come out below one. Those issuers are judged instead on the issue’s own record, measured from its listing date: ten years or more of payments earns full credit, five years scores 7, two years 5, anything younger 3.
The distinction matters more than it sounds. Under the previous version, one of the largest banks in the world scored zero on the heaviest measure, and the whole mortgage-REIT complex with it, not because those preferreds were weak, but because the arithmetic was the wrong arithmetic for that kind of company.
Credit
The second measure reads the financial strength of the company behind the preferred. The most reliable gauge an individual investor can get is the credit rating.
The score grades the full rating ladder point by point, so a stronger rating always counts for more than a weaker one: the top of the ladder and the last investment-grade step are both investment grade, but they are not the same claim. Where Moody’s and S&P disagree, the score takes the more cautious of the two, as the index providers do. A quality score should take the careful side of a disagreement; the optimistic side is the case the issue’s marketing has already made.
Roughly half the preferred market carries no Moody’s or S&P rating at all. For those issues an Egan-Jones rating fills in, counted one step below its face reading as a cautious adjustment, because its ratings are paid for by investors rather than by the companies rated and its scale does not line up exactly with the other two. An issue rated by none of the three goes unscored on this measure, and the others pick up its weight. An unrated issue is never quietly treated as a zero.
One caveat deserves plain statement. Rating firms usually grade a preferred one to three steps below the same company’s regular debt, because preferred holders stand further back in line if things go wrong. A preferred at the bottom of investment grade often comes from a company whose bonds sit several steps higher. The IQS uses the preferred’s own rating, because that is the rating that governs what the investor actually holds.
Buyback economics
Most preferreds give the company the right to buy them back at face value, usually starting five years after they are sold. That right matters, because a buyback ends the income on the company’s schedule rather than the holder’s.
Version 3 weighs the calendar and the price together, because the date alone does not decide anything. Five or more years of protection earns full credit, stepping down as the date approaches, and a preferred the company can never buy back stays at 10. Price then adjusts the reading: shares trading well above face value lose up to three points, because the holder has something real to lose in a redemption, while shares trading well below it gain two.
Once the date has passed, price decides outright. An issue at a deep discount scores 9: a buyback there would hand its holder a gain, which is not a risk. One near face value scores 6. One above face value scores 3, because the company can refinance profitably and probably will. The previous version scored every past-date issue zero regardless of price, which treated a windfall and a threat as the same thing.
Structure
A preferred dividend is not a bond coupon. The company can put it off or stop it altogether, and what happens next depends on the fine print. On a cumulative preferred, skipped payments pile up as a debt the company must clear before it can pay its common shareholders again. On a non-cumulative preferred, a skipped payment is gone for good.
Version 3 judges that against each industry’s own conventions rather than as a flat yes-or-no. A cumulative preferred earns 10. A baby bond earns 9, because its interest is a contractual obligation and it ranks ahead of preferred shares. A bank preferred without the make-good promise earns 8; since the financial crisis, regulation requires a bank preferred to drop that promise for the shares to count toward the bank’s capital cushion, so following the rule is not a defect. Outside the financial sector, where 99% of issues do carry the promise, an issue without it is the genuine exception and earns 5 with a flag for review.
Protection
The newest measure is the oldest test on the preferred desk: is the company still paying a dividend on its common stock?
The answer matters because of the order of the queue. A company must stop paying its common shareholders before it can stop paying its preferred ones. So a common dividend still being paid is a live, checkable signal that the preferred sits behind a cushion, and it scores 9. A company that has suspended its common dividend scores 1, with a red flag, because that cushion is gone. A company that never paid a common dividend scores 3: there is simply no cushion, which is a different thing from distress.
Penalties, missing data and confidence
The weighted average of the five measures is the starting score. Certain events then cut it down directly. A recent dividend cut or suspension multiplies the score by 0.55. A likely deferral, a make-good issue that has gone more than two payment cycles without declaring and is not already marked suspended, multiplies it by 0.45. Two further checks cut the score by 15% each: a price below $10 on a $25 issue, and a company whose debts exceed everything it owns. All can apply at once, the score can never fall below a fifth of its starting value, and each cut shows up in the product as a red flag next to the number, so the deduction is never hidden. An issue the company has already redeemed stops being scored.
When one of the five measures cannot be scored, because there is no usable earnings figure and no payment history, or no Moody’s or S&P rating, the other measures pick up its share of the weight. The product reports how much of the data was actually available and grades its own confidence next to the score: high when 85% or more of the weight was scored, medium at 60%, low below that. A 9.0 built from complete information and a 9.0 built from half of it are different claims, and the product says which one is on the page.
The score also carries a plain-word label wherever a word helps more than a number: High Quality at 8.0 and above, Good from 6.5, Average from 5.0, and Watch below that.
How the measures interact
The five measures are scored separately, but they tell on each other, and reading them together is where the score earns its keep.
The clearest pairing is Coverage and Credit. Coverage is arithmetic: how far current earnings clear the payment. The rating is a judgment about whether the company can keep paying through a bad stretch. Strong earnings under a weak rating often mean a company doing well today with longer-term problems the rating firms can see. A strong rating over thinning earnings can be an early warning the firms have not yet acted on. The measure-by-measure breakdown shows this; the total score alone hides it.
A second pairing connects Buyback economics with Coverage. The buyback measure already reads price and date together, so a low score there means a redemption would genuinely cost the holder something. Whether that matters depends on how well the payment is earned in the first place: a richly covered issue trading above face value is a good problem, while a thinly covered one is two risks pointing the same way.
A third pairing is Structure and Credit under stress. Structure says what the investor is owed if payments stop. Credit and coverage say how likely payments are to stop in the first place. A cumulative preferred from a shaky company may well need its protection; a non-cumulative preferred from a strong company may never test its weakness. The score keeps the measures separate because they answer different questions, but a careful reader holds both in mind.
Scoring BDCs
A business development company is not a preferred stock. It is a lender, and its shares are ordinary stock in that lending business. The board sets the dividend each quarter; nothing is promised in advance, there is no face value, and there is no buyback date. So the questions change. Is the dividend actually earned by the lending? How much borrowed money sits behind it? What is the market paying for a dollar of the loan book? And how much do the managers take for themselves?
The BDC score follows the same rules as the preferred score: measures graded in steps, a 0-to-10 result, weight shifted to what could be measured when something is missing, the same penalties, labels and confidence grades. Seven inputs feed the five measures shown in the product:
NII coverageis the anchor, and it plays the role earnings coverage plays for preferreds. It compares the company’s lending income per share with the dividend it declares, giving full credit when income runs 1.6 times the dividend and none at 0.8 times. A BDC paying out more than its loans earn is funding the dividend some other way, by selling assets, handing back capital, or issuing new shares, and none of those last. Market carries the largest displayed weight because it reads the two prices the market sets: the yield, graded from 5% up toward 15%, and the share price against the stated value of the loan book. Full credit goes to shares trading between 90 cents and $1.05 per dollar of book value, where the market neither doubts the loans nor pays a premium that eats into future returns. Credit fades over the next twenty cents in either direction.
Leveragegrades borrowed money against the company’s own capital, from half a dollar of debt per dollar of equity down to two dollars, which is roughly where regulation stops a BDC anyway. Beyond that line the overall score takes a further penalty. Cost & scalecombines what the managers charge, the base fee plus a tenth of the performance fee, graded from 1.5% to 3.5%, with the company’s size, from $200 million toward $5 billion. Small BDCs tend to carry both higher costs and thinner trading. Portfolio quality grades how much of the loan book sits first in line to be repaid, from 30% to 90%. When borrowers run into trouble, that ordering decides how much of the book survives.
Two events cut the score directly. Debt above the two-dollar line multiplies it by 0.85. Lending income below the dividend, meaning a payout not currently earned, multiplies it by 0.8. Both appear as red flags beside the score.
Scoring preferred ETFs
A fund that owns preferreds is judged on what it holds and what it charges to hold it, under the same rules as everything else the score touches. The measures: portfolio quality at 40%, expenses at 20%, earned payout at 15%, breadth at 15%, and size at 10%.
Portfolio qualityreads the credit breakdown the fund’s own manager publishes each month, the share of the book rated investment grade, rated below it and unrated, converted to the same scale the preferred credit measure uses: investment grade counts as 10, rated high-yield as 4.5, and unrated as a neutral 5, because half the preferred market carries no rating and unrated is not the same as junk. The manager’s figure leads rather than our own records because these funds own $1,000 institutional preferreds that no listing of exchange-traded holdings can cover. Our own records serve as a cross-check when at least six of a fund’s holdings match our coverage, and a gap of more than 2.5 points between the two sources is flagged for review rather than papered over.
Earned payoutasks whether the fund earns what it pays, rather than rewarding whichever fund pays the most. The payout rate is compared with what the portfolio itself earns after expenses. Paying out up to 1.3 times what the book earns takes full credit, since a portfolio bought at a discount legitimately yields somewhat more than its stated rates. Beyond that the grade falls quickly, 7 at one and a half times, 4 at 1.8, 2 at 2.2 and zero above, because a payout far past what the book earns is borrowed money or the investor’s own capital coming back.
Breadth grades the number of holdings the fund itself reports, from full credit at 200 or more down to 2 below twenty. Expenses grade from 0.10% of assets at full credit to 0.80% at none, and size from $100 million toward $5 billion.
Two protections sit on top. A fund that borrows to amplify its returns, or that is structured as a single bank’s note rather than a fund holding assets, is cut by a fifth and flagged. And a fund with neither a published credit breakdown nor a usable cross-check from our records can never score above 6.9, the top of the Good range, because missing data must never read as excellence.
What changed at version 3
Version 3 replaced the preferred and preferred-ETF scores on September 1, 2026, after computing silently alongside the previous version across the whole market. Four changes drove it, each correcting a way the old score misread the market rather than the securities.
Banks and mortgage companies were scored near zero by construction. The old coverage measure divided operating profit by interest cost. For an industrial company that is a fair test. For a bank, paying interest on deposits is the business, so the ratio comes out below one for even the soundest names, and the grading turned that into a zero on the heaviest measure. Mortgage REITs, funded the same way, were caught by the same defect. Version 3 routes those issuers to a different test.
The old score took the more flattering of two credit ratingsand collapsed everything investment grade onto a single top mark, so the strongest rating on the scale and the weakest investment-grade one were indistinguishable. Version 3 takes the more cautious view, grades the full ladder point by point, and reads Egan-Jones where Moody’s and S&P are both silent, which is roughly half the market.
Bank preferreds were penalized for following the rules.Since the financial crisis a bank preferred must drop the make-good promise for the shares to count toward the bank’s capital. The old structure measure read that as a defect. Version 3 judges the fine print against each industry’s own conventions.
The buyback measure ignored price. An issue past its buyback date trading well below face value has nothing to fear from a redemption, it would hand the holder a gain, yet it scored the same zero as one trading above face value that the company could profitably refinance. Version 3 weighs price and time together.
The safeguards tightened too. A score built on less than half its data is pulled toward a neutral 5 rather than being allowed to look impressive on two facts. Implausible figures, a balance sheet off by a factor of a thousand, a price under $10 on a $25 issue, a company whose debts exceed everything it owns, are screened or penalized instead of scored at face value. And thin trading became a disclosure beside the score rather than a deduction inside it, because a preferred is not a trading instrument and a quiet issue that has paid for decades is not lower quality for being quiet.
Because these were corrections rather than news about any security, the changeover produced a one-time step in every affected score on the day it landed, and nothing else. Business development companies and covered-call funds were not part of this revision and keep the methods described above.
Scoring covered-call ETFs
A covered-call fund is a different bargain again. The investor puts up capital and gives away most of the upside; the fund pays a large monthly distribution in exchange. The advertised rate is the least useful number on the page. What matters is where the money comes from: option income and gains, or the investor’s own capital handed back. The tax rules make that surprisingly hard to see, because the same “return of capital” label sits on both a fund paying out a shrinking portfolio and a healthy index-option fund whose payouts are simply taxed that way while the portfolio grows.
The covered-call score follows the same rules as the others: measures graded in steps, a 0-to-10 result, weight shifted to what could be measured, and the same refusal to show a score built on too little data. The weights themselves were set by testing, not debate. Every candidate measure was scored using only what an investor could have known at a year-end, then checked against the return that followed, and the weights follow that evidence. Tested results describe one stretch of history and promise nothing about the next.
Payout integrityis the anchor. It measures how much of the distribution is the investor’s own capital coming back, always from a documented source, never a label. The final year-end tax form comes first. When that is not yet out, the fund’s own running estimate is used, and it is shown as an estimate, since these tend to run high. For funds whose fiscal years leave no calendar-year figure, the audited annual report fills in. Two checks against the fund’s one-year return then apply. If the fund lost money, this measure can score no better than 3, because paying distributions out of a shrinking portfolio is exactly the failure the score exists to catch. If the fund returned 8% or better, it scores at least 5, because a high return-of-capital share beside a growing portfolio is a tax outcome, not lost capital, and the score does not treat it as damage.
Exposurecarries equal weight because it showed the strongest link to later returns in our testing. It grades what the fund sells options against. A broad index earns full credit, and the grade steps down through bond, basket and sector funds to a single stock at zero. Two points come off where the fund bets against the market, where the payout is owed by a single institution rather than an options market, and where the fund’s exposure runs past 100% on borrowed money. Capital preservation grades the one-year return with distributions counted as reinvested. Funds between six and twelve months old are judged on the year so far, and younger funds leave the measure blank rather than guessed at.
Cost grades the full expense rate from the prospectus fee table, from 0.35% at full credit to 1.50% at none. Where a fee waiver with a stated end date is in force, the lower rate counts until it expires. Scale & continuity asks whether the fund will still be here in a few years: verified assets graded from $10 million toward $1 billion, with the sponsor’s figure checked against regulatory filings before it counts, and age caps that hold any fund under six months to 4 and under a year to 6 no matter its size. A large young fund still has no record.
The same restraint applies when data is missing. When less than 60% of the required information is available, no score is shown at all, rather than a low number a reader might mistake for a verdict, and a fund that has never documented where its payout comes from is never marked better than partly measured, whatever else was available. Under these measures, as of August 2026, 143 of the 155 funds in the covered-call universe receive a score. The twelve that do not are recent launches with no return history. The earlier framework, which relied on figures no sponsor publishes, could not bring a single fund to full coverage.
Refresh cadence and event-driven updates
The scores are recomputed every day, but not every measure moves every day. Each one updates on the schedule of the data behind it.
Coverage updates when the company reports earnings or changes its dividend. The inputs come from quarterly filings and declaration notices, so the measure moves on the reporting calendar. Credit updates whenever a registered rating firm changes its view of a company we follow. Rating actions arrive through licensed market-data feeds and SEC filings, and the score reflects them at the next recompute, every trading day after the close. Structure updates whenever a company declares, suspends or reinstates a preferred dividend. Routine declarations change nothing; suspensions and reinstatements move the score at once. Call protection updates daily by the calendar, easing down as the buyback date draws closer.
Covered-call measures move on the tax calendar. Payout integrityresets each January through March, when sponsors publish final tax figures for the prior year. Between year-ends it runs on the funds’ own estimates, shown as estimates, and the yearly reset is itself a score change. Cost steps up from the discounted rate to the full one when a fee waiver expires, and scale & continuity updates as sponsor asset figures are refreshed and checked against quarterly filings.
When a measure moves enough to shift the overall score by a point or more, the move is visible on Market Trends, which lists the week’s largest score changes in each class, so a reader can see not only where a security stands but how it has moved lately, which is often as telling as the level itself.
The IQS in practice
The example below is an illustration. It reconstructs a score card for Gladstone Investment’s 7.125% Notes Due 2031, a real baby bond that traded in the US over-the-counter markets, using figures as of mid-2026. It is not a live score, and the current one may differ. The inputs are the kind the system draws from SEC filings, licensed market data and Egan-Jones credit ratings, with no reference to anyone else’s scoring method.
GAING comes out at 6.8, which carries the Good label. No penalty applies. The number is simply the weighted blend, and every lost point has a visible cause. A different security could arrive at the same 6.8 with its weak spots in different places, and what that means for the investor would be different too. The score is identical; the shape is not.
Walking through the measures makes this concrete. Gladstone Investment is an operating company, not a bank, so coverageruns on the earnings test: pretax cash flow covers the notes’ interest about 2.4 times over, a real cushion but not the five-times margin that earns full credit. Creditis where version 3 changes this card. GAING carries no Moody’s or S&P rating, which under the previous version left the measure unscored and the score published on 73% of its weight. An Egan-Jones rating now fills that gap, counted one step lower as a cautious adjustment, so the score is built on all five measures. Buyback economics sits in the two-to-three-year range with the May 2028 date 24 months away, and the notes trade near face value, so price neither helps nor hurts. Structure earns 9: interest on a baby bond is an obligation rather than a choice, and it ranks ahead of preferred shares. Protectionearns 9 because Gladstone still pays its common distribution, which must stop before the notes’ interest can. The blend lands at 6.8.
The breakdown travels with the score wherever it appears in the product. An investor scanning a watchlist sees the single number for quick comparison. The same investor, opening a holding, sees the measures that explain why the number is what it is. Both forms are always available; neither replaces the other.
Not investment advice. The Income Quality Score is an analytical framework intended to assist income investors in evaluating preferred securities. It does not constitute investment advice, a recommendation to buy or sell any security, or a guarantee of any particular investment outcome. Investors should consult their own financial adviser before making investment decisions.
How to read a score
The score is designed to be read, not obeyed. Two habits of reading get the most from it.
Where a score points.The score is not a substitute for reading the prospectus; what it does is surface which sections deserve the closest attention. A preferred scoring 7 with the deduction in Call protection directs the reader to the call-provisions section; a deduction in Coverage directs them to the issuer’s latest earnings. The number is a table of contents for the diligence, not a conclusion from it.
What the shape shows. Two preferreds with the same overall score can have opposite profiles. One is strong on coverage and credit but close to its buyback date; the other is the reverse. The breakdown makes that difference visible instead of collapsing it into a single ranking. Which profile matters more is a question the score raises and does not answer.
One thing the score never offers is advice on how much to hold. It describes the quality of a single issue; it says nothing about what share of a portfolio that issue should be. That depends on the investor’s overall holdings, taxes, income needs and appetite for risk, none of which the score can see.
What the IQS does not attempt
Three things are worth stating plainly, because scoring systems are often assumed to do them.
The IQS is not a recommendation. A preferred with a high score is not automatically a good purchase at today’s price. A preferred with a low score may be a perfectly sensible holding for an investor whose situation calls for more yield in exchange for more complexity. The score describes the security; it does not advise on the trade.
That principle extends to the product built around the score, and two rules follow. First, the score never drives notifications. YieldDesk alerts fire on things that happen at the security itself: a price level the user sets, a declared distribution, a call notice, an insider filing. They never fire on a movement in the score. Second, the score is never combined with a reader’s holdings to produce rankings or personalized output. A security’s score is the same for every reader and knows nothing of anyone’s positions. Holdings pages show the same figure every other page shows, and nothing more.
The IQS does not predict returns. We make no claim that high-scoring preferreds beat low-scoring ones over any holding period. The score measures qualities we believe matter to income investors making long-term decisions. Whether the market has already priced those qualities in at any given moment is a separate question.
And the IQS is not a substitute for reading the prospectus. The five measures are numerical summaries of complicated legal terms. The prospectus holds the full text, from the buyback provisions to the payment mechanics to any conversion features, and an investor putting real money into a single issue should read it. YieldDesk links the issuer’s EDGAR filings from every record.
Looking forward
The IQS continues to evolve. The measures reflect our current best judgment about what matters most to investors who hold these securities one at a time, and the shift to a coverage-anchored design from the earlier framework was exactly that kind of refinement. As markets change, as new structures appear and as readers tell us what helps, the method will keep improving. Every change is documented on this page, with a version number, a date and a plain account of what changed and why. The current version is 3.0, of September 1, 2026. It rebuilt the preferred and preferred-ETF scores: coverage judged by how the company makes its money, credit graded on the full ladder and on the more cautious of two ratings, buyback economics reading price alongside the date, fine print judged against each industry’s conventions, and a new protection measure on the common-dividend test. Business development companies and covered-call funds were not part of it. Version 2.1, of August 21, 2026, added the covered-call framework, five measures with weights set by testing against history in place of an earlier set that leaned on figures no sponsor publishes, and revised the reading guidance so the page describes what a score shows rather than what to do about it. Version 2.0, of August 12, 2026, spelled out the step-by-step grading, introduced the penalties, reweighting and confidence grades described above, and added the BDC framework.
The score runs through the whole product. The full breakdown appears on every record; the single number appears on the watchlist, in the screener and wherever space is tight; and Market Trends lists the week’s largest moves. We will keep building the product around it, and keep publishing changes to the method here.