Grant the claim first, because most of it is reasonable. A company with real margins and little debt is not fighting for its life, and a company fighting for its life tends to have a share price that jumps around. The link is intuitive and it is partly real.
We wanted to know how much of it holds once you account for the most obvious confounder.
The first pass looked like a clean sweep
Turbulence here is the rolling variance of daily returns, annualised. It measures the range of a company’s movement, not its direction. A high reading means the stock has been covering a lot of ground.
We sorted the panel by twelve measures. For each one we took the median 90-day turbulence of the top fifth and divided it by the bottom fifth. A ratio below 1.00 means the high end of that measure was calmer.
| measure | ratio | measure | ratio | |
|---|---|---|---|---|
| Market cap | 0.17 | Net margin % | 0.24 | |
| Price vs 52w range | 0.23 | Interest coverage | 0.29 | |
| ROIC % | 0.24 | RSI 14 | 0.35 | |
| Piotroski F | 0.28 | FCF margin % | 0.37 | |
| Altman Z | 0.50 | Debt / equity | 0.67 | |
| Revenue growth YoY | 0.97 | Current ratio | 3.16 |
Ten of the twelve point the same way. Higher return on capital, fatter margins, better interest coverage, a stronger Altman Z, a higher Piotroski score: all of them line up with calmer stocks. On this evidence the claim looks settled.
Two of the twelve are not measures of quality at all. Price against the 52-week range and RSI describe what the price has been doing, so they ought to track turbulence, and they are on the list as a check that the test can find something when something is there. They do.
Then look at the first number in the table.
Size is doing most of the work
Market capitalisation separates turbulence better than any measure of quality on the list. The largest fifth of the panel has a median turbulence of 0.095 against 0.553 for the smallest fifth, roughly six times calmer, and nothing else comes close.
That matters because the quality measures are not independent of size. Unprofitable companies with weak balance sheets and low Altman Z scores are disproportionately small. So a table showing that low-quality companies are turbulent may be showing something much simpler: small companies are turbulent, and small companies are more often the ones with thin margins.
Twelve findings, or one finding wearing twelve costumes. There is a standard way to tell the difference.
The size-controlled re-run
We split the panel into ten market-capitalisation deciles and ran the same comparison inside each one, top third against bottom third. Within a decile every company is roughly the same size, so size can no longer explain the result. We also recorded how many of the ten deciles pointed the same way, because a measure that works in five deciles and reverses in the other five is noise with an average.
| measure | whole panel | within deciles | consistent |
|---|---|---|---|
| Net margin % | 0.24 | 0.48 | 10/10 |
| Price vs 52w range | 0.23 | 0.48 | 10/10 |
| ROIC % | 0.24 | 0.52 | 8/10 |
| Interest coverage | 0.29 | 0.53 | 8/10 |
| RSI 14 | 0.35 | 0.60 | 9/10 |
| FCF margin % | 0.37 | 0.73 | 9/10 |
| Piotroski F | 0.28 | 0.82 | 9/10 |
| Altman Z | 0.50 | 0.89 | 5/10 |
| Debt / equity | 0.67 | 0.99 | 5/10 |
| Revenue growth YoY | 0.97 | 1.40 | 9/10 |
| Current ratio | 3.16 | 1.67 | 9/10 |
Every single measure moves toward 1.00. That is the finding in one sentence: a large part of what looked like quality was market capitalisation answering in its place.
Three groups came out of the other side.
Profitability survives. Net margin holds at 0.48 and points the same way in all ten deciles. Compare two companies of the same size and the one with the better margin has been the calmer stock, reliably. Return on capital and interest coverage are weaker versions of the same result at 8 deciles out of 10, and free cash flow margin at 9.
Piotroski F sits with this group rather than with the balance-sheet measures, at 0.82 across 9 of 10 deciles. It is the weakest survivor on the list, and it is a survivor.
Two balance-sheet measures dissolve. Altman Z weakens from 0.50 to 0.89 and points the same way in only five of the ten deciles, which is what a coin does. Debt to equity lands at 0.99 on five of ten, which is the cleanest null result on the page. Whatever those two were tracking in the first table, most of it was size.
Altman Z is one of the two lenses this company was built on, so that result is worth stating plainly rather than burying. It is also narrower than it looks, which is the next section.
Two measures run backwards. Companies with more short-term liquidity have been more turbulent, not less, at 1.67 across nine of ten deciles. So have faster-growing companies, at 1.40 across nine of ten. Both are consistent enough to take seriously. A high current ratio often describes a company holding a large cash pile against very little revenue, and the market moves those around a lot. Fast revenue growth is priced on expectations, and expectations move.
What this does not say
It does not say the Altman Z-score is broken. This tested one thing: whether a company’s Z-score tracks how much its stock moves. That is not what the Z-score is for. It was built to identify companies heading for distress, and our own failure work continues to find that it does that job. A measure can be poor at explaining share-price movement and good at flagging distress, and Z appears to be exactly that. The honest version of this finding is narrow: the appearance that balance-sheet quality tracks market calm is mostly market capitalisation.
It is one cross-section, not a history. Every number here comes from the panel as it stood on 2026-08-04, with turbulence measured over the preceding 90 days. It describes a single moment in a single market. A multi-year version of this test is a different piece and we have not run it.
The twelve measures were a scan, not a prediction. We did not name in advance which ones would survive. The size control was fixed before we looked, and the ten-decile consistency count is there to make accidental results harder to keep, but this is exploratory work and should be read that way.
Deciles are a coarse control. Companies inside one decile are similar in size, not identical, so some size effect can survive the split. A continuous control would be better and we have not run one.
It is not advice. Nothing here says anything about what any company is worth or what anyone should do about it.
Method
Panel of 2,662 US-listed companies as of 2026-08-04: every row carrying both a turbulence
reading and a market capitalisation. Market capitalisations come from the promoted
universe file, not the panel’s market_cap_m field, which has a known extraction defect.
Turbulence is the 90-day rolling variance of daily log returns, annualised on 252 days.
Comparisons use medians throughout, so a handful of extreme names cannot carry a result.
Deciles are formed on market capitalisation within the same panel date.
One measure needed different handling. Piotroski F is a whole number from 0 to 9, and 439 companies share a single value, so a cut at the top or bottom fifth by position lands inside a block of ties and the answer moves with the sort order. Discrete measures are therefore cut by value rather than by position: every company at or below the twentieth-percentile score against every company at or above the eightieth. Doing it the other way changes Piotroski’s reading materially and does not change any other measure on the list.
Descriptive only. Not advice, not a recommendation, not a forecast. Turbulence measures the range of a company’s movement, not its direction.