Understanding accruals

[yuvaltaylor]

The accrual ratio that I’ve been using, to extremely positive effect in my backtests, is (NetIncBXorTTM-OperCashFlTTM)/AstTotTTM, with lower values being better. Using this formula heavily (about 10% or 12% of rank) instead of certain other “quality” measures improves my excess returns by up to 20% over any other system I’ve tried. But I don’t quite understand why. I suspect that it’s because it’s opening my portfolio up to growing companies with negative earnings rather than because it has any merit on its own—I’ve coupled it with a heavy emphasis on growth (earnings, operating income, revenue) and value in terms of price-to-sales. In terms of measuring quality, a reliance on ROE and ROCE will effectively eliminate companies with negative earnings, while the accrual ratio actually favors such companies.

Take two scenarios, for example. You have a company that’s declining. Your income is low, and prospects don’t look good. So you mine your inventory and receivables to generate additional cash. Your operating cash flow will be terrific, your net income will be low, and your accrual ratio will be negative–i.e. superb. But your company is in deep trouble. Or take the opposite situation. Your company is growing like crazy. You’re making a nice, tidy profit. And you’re investing your extra cash in additional inventory and accounts receivables, resulting in a negative net cash flow. So your accrual ratio is super high, an indication of earnings manipulation and danger ahead, but your company is going to do just fine.

Where am I going wrong here? I believe I’m making a fundamental mistake somewhere, but I don’t know what it is. I know that Sloan demonstrated that companies with low or negative accrual ratios really outperform those with high accrual ratios, but it doesn’t make sense to me given those two scenarios I just gave.

[mgerstein]

I think the key to both of your examples lies in the fact that NetIncBXorTTM is a large component of OperCashFlTTM.

Your hypothetical deteriorating company not only has declining income but usually declining operating cash flow. When deteriorating companies mine inventories and receivables to generate cash chances are its at a level that is not sufficient to allow their income to turn upward (maybe they are slashing price to clear merchandise that can’t be sold at profitable prices).

Meanwhile, if your growing company is using its cash well, we should presume income will benefit, if not immediately but within the realm of reasonable expectation.

Not sure which Sloan paper you read but he and the other earnings quality folks are not moralists. Manipulation etc. would be fine with them expect for the poor association between manipulation and persistency. High accruals are troublesome because in the aggregate, they presage less-persistent earnings streams (Enron being the penultimate poster child).

The reason why TATA (total accruals to total assets) is associated with lower persistence is because accruals come from human choices and that inherently raises questions of reliability that are not present with the cash component of earnings. This is not to say that all such judgments are bad. The goal of GAAP earnings is to match revenues with the expenses necessary to generate them, and that can’t be done without discretion and accruals because revenues in every period incur at least some expanses that were incurred in different periods. So accruals are not inherently evil and we cannot ignore Net Income. But we have to weigh and balance that with the message of Operating Cash flow.

So one thing you are likely seeing is the impact of aggregation. While on the whole, accruals are legit but and in some cases desirable, your statistical work is aiming at identifying the less functional aspects of the universe.

There’s another angle. All else being equal, less uncertainty is preferred and if we assume that complexity and uncertainty are correlated, then we could say that simplicity is preferred. The more accrual heavy an income statement, the further away you get from the most basic accounting concepts and the more readily you are called upon to address more substantial accounting ideas; i.e. complexity which equals uncertainty.

All that said, there is interesting research out there that decomposes accruals into different types, those that are inherently more or less reliable, and even strays beyond the current portions of the balance sheet accounts.

[yuvaltaylor]

Thank you, Marc. That really helps explain things.

I have a somewhat related question. There are companies that are well worth investing in whose operating income and net income are both negative. Obviously, ROE and return on capital are going to be useless for measuring quality for those companies. Do you think the accrual ratio is helpful for judging companies such as these? Do these companies really fit in with the DDM/DCF model? Are there other quality ratios that are consonant with the DDM approach that might substitute for ROE or return on capital? For valuation, are the only choices the price-to-sales and price-to-book ratios?

[mgerstein]

Great question!

There are lots of things we can do with companies like that, and we should always resist the temptation to allow ourselves to be lulled into the idea that rational fundamental analysis requires us to go after goody-two-shows companies. Excellent returns can be achieved from shares of putrid companies that look like they are or may become a bit less stinky.

ROE etc. can be fine but you would need different expectations. You would be looking here for improvement in ROE etc. over time and/or relative to a peer group. As nice as profits are, one should never underestimate the way a stock can be stimulated by horrible losses that improve to losses that are merely bad, or the way losses transitioning to profits can be beneficial.

TATA and/or Beneish M Score could help too (impacting the risk component of R) because companies that seem weak might have more incentive to account aggressively and because there may be fewer eyeballs looking. It’s not so much a matter of favorable M Scores or low TATA as it is eliminating the worst of the worst (screening rules might be more effective than rank factors in this regard).

For valuation purposes, you’d think in terms of a sales-based DDM variation.

P/S or EV/S = Margin/(R-G)

Look for growth in sales

Look for improvement in margins (not just net, look at gross, operating and/or ebit margins)

Look for ROE trending in the right direction as a cue that the risk component or R might be favorably reassessed down the road

Look at financial strength (debt ratios, cash flows); ,many companies seemingly weak companies have stronger balance sheets than Mr. Market might expect because their CFOs are more concerned with meeting payroll than buying back stock so this could make the risk component of R lower than the market presumes

Finally, look for rising noise (technical, momentum, sentiment factors) especially among stocks where noise is low to begin with. The Street can easily get over-excited about even less-substantial positive news.

[davidbv]

I learned something from Marc’s explanation above, as I usually do when he comments.

Could another take on this be:
If you are minimizing (NetIncBXorTTM-OperCashFlTTM)/AstTotTTM, then that means you are looking for NetInc to roughly equal OperCashFl.
For this to happen, Depr&Amort+other has to be about zero.
Assuming that ‘other’ is consistently a small part of a companies cash flow statement (on a regular basis), then that implies Depr&Amort is close to zero.
For Depr&Amort to be close to zero, that implies that a companies capex is very low or zero. Good companies, on average, should have their Depr&Amort roughly equal their capex (assuming they have little intangibles to write off). This insures they are reinvesting to replace their physical assets as it ages.
So maybe this equation of yours is zeroing in on companies that have little capex. There are many examples of good companies who have little capex that have done very well since 1999. In fact, maybe most of them.
I was just reading a thought from Buffett that a wonderful business is one that creates profit from as little capital investment as possible. I guess that would include capex (in addition to equity and debt - hopefully I am interpreting his comment correctly). I know many investment managers (Peter lynch comes to mind) that looked for non-capex intensive companies too.

[yuvaltaylor]

Actually, you’re looking for OperCashFl to substantially exceed NetIncBXorTTM. A lot of energy companies these days are in this situation, with negative incomes and positive cash flows. Most of these companies invest heavily in Cap Ex. And depreciation and amortization expenses are deducted from net income but ADDED back to operating cash flow. So a company with a low accrual ratio is actually MORE likely to have substantial capital expenditure spending. For example, look at Freeport-McMoran. A TTM loss of 14 B and operating cash flow of over 3 B, creating an accrual ratio of -0.22. Why? A good part of the reason is Cap Ex of 5.5 B and Dep&Amort of 3.5 B.

[davidbv]

Ach. The more I think about it, I see what you are saying. I was thinking about it in the context of taking the absolute value of the difference (which would minimize it). You are truly looking at the difference. Good point. Interesting.

[yuvaltaylor]

No, my ranking system gives very high ranks to companies like REGI and KOP which have TTM net incomes well below 0 and healthy cash flows. The huge losses these companies have incurred made me nervous. But Marc has made me understand that the combination of high growth, negative income, and healthy cash flow can make for a good investment opportunity.

[davidbv]

yes, I see it now. Good point. Thanks.

[ericsull]

@yuvaltaylor

No, my ranking system gives very high ranks to companies like REGI and KOP which have TTM net incomes well below 0 and healthy cash flows. The huge losses these companies have incurred made me nervous. But Marc has made me understand that the combination of high growth, negative income, and healthy cash flow can make for a good investment opportunity.

Hi Yuval,

Do you use the absolute value of the negative earnings in your formula (NetIncBXorTTM-OperCashFlTTM)/AstTotTTM if the company has posted a loss?

A company with a $50 loss and a $50 OCF would have the same accruals as a company with a $50 profit and $150 in OCF, assuming assets are the same. Would that be correct?

Thanks,
Eric

[yuvaltaylor]

@ericsull

Do you use the absolute value of the negative earnings in your formula (NetIncBXorTTM-OperCashFlTTM)/AstTotTTM if the company has posted a loss?

No, I do not.

@ericsull

A company with a $50 loss and a $50 OCF would have the same accruals as a company with a $50 profit and $150 in OCF, assuming assets are the same. Would that be correct?

Yes, that's correct.

Here's ChatGPT on the subject, in case it helps you think about it:

When measuring accruals as the difference between net income and operating cash flow, accruals represent the adjustments made to net income to account for non-cash items and changes in working capital that affect cash flow. The calculation of accruals in this context typically includes the following components:

1. Changes in Working Capital:

   • Accounts Receivable: Increases in accounts receivable decrease cash flow but increase net income, so this would be added back to accruals.
   • Inventory: Increases in inventory reduce cash flow but don’t immediately affect net income, so this is added to accruals.
   • Accounts Payable: Increases in accounts payable increase cash flow but don’t affect net income immediately, so this would be subtracted from accruals.
   • Other Current Liabilities/Assets: Similar adjustments for other current liabilities (e.g., accrued expenses) and other current assets that affect cash flow but not net income immediately.

2. Non-Cash Expenses/Income:

   • Depreciation and Amortization: These are non-cash expenses that reduce net income but do not affect operating cash flow, so they are part of accruals.
   • Deferred Taxes: These represent tax expenses recognized in net income but not yet paid in cash, thus included in accruals.
   • Provisions and Reserves: These are expenses accrued in net income but not yet paid in cash (e.g., warranty reserves or bad debt provisions).

3. Non-Operating Gains/Losses:

   • Gains or Losses on Asset Sales: These affect net income but are considered investing activities in cash flow, so they are part of accruals.

4. Other Adjustments:

   • Unrealized Gains/Losses: Changes in the fair value of financial instruments or foreign currency translations that affect net income but not operating cash flow.

In essence, accruals include all the items that adjust net income to reconcile it with the operating cash flow, covering changes in working capital, non-cash expenses, and other adjustments that impact net income but do not directly affect cash flow during the period.

[ZGWZ]

I don't think accruals are necessary where metrics about FCF/EBITDA exist. However, there may be some benefit to throwing the accrual factors into the AI system.

Because of the weak international evidence for accrual-type factors, I think they may have to be given slightly less weight.

[yuvaltaylor]

@ZGWZ

Because of the weak international evidence for accrual-type factors, I think they may have to be given slightly less weight.

Out of the best five factors in the US market over the 1975-2019 period, two are accruals-related, according to https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3774514

• cash-based operating profits to total assets
• one-year current operating working capital change to total assets

A large number of the best-performing factors that the authors tested are variations on one central concept: that net operating assets, cash-based accruals, and net operating working capital, which are all closely related, should be low in comparison to total assets.

[ZGWZ]

That's what I meant:

Also, it is not so good in US microcap universe:

[RTNL]

Where is that study from please?

[ZGWZ]

It is from Yuval's pdf

[ScifoSpace]

@yuvaltaylor

(NetIncBXorTTM-OperCashFlTTM)/AstTotTTM

Accruals-to-Price Not Working — Am I Missing Something?

I’ve been experimenting with this formula:

CopiarEditar

(NetIncBXorTTM - OperCashFlTTM) / AstTotTTM

It’s meant to proxy accruals relative to total assets — a variation of the classic accruals-based quality signal. Inspired by What Works on Wall Street, I also tested Accruals / Price (with lower being better), as it's highlighted in the book as one of the stronger predictors of future returns.

However, in my own backtests, the standalone performance of this metric has been quite poor. Results are often noisy, weak, or even inversely correlated with what’s expected, especially across broader universes.

So I’m wondering:

• Is anyone here using accrual-based indicators successfully in practice?
• Are there specific universes (e.g., small caps, low analyst coverage, non-financials) where this performs better?
• Does scaling by net income, assets, or using longer time horizons (e.g., multi-year medians) improve the robustness?
• Or maybe this metric just needs to be combined with other factors (like value or profitability) to really shine?

Curious if @Yuvaltaylor or anyone else with experience in this area has insights. Would appreciate any ideas, corrections, or examples. I feel a bit lost on whether this signal is fundamentally broken — or if I’m just using it in the wrong context.

Thanks in advance!

[yuvaltaylor]

The accrual ratio is an important tool for determining whether a company may be manipulating its earnings. I use it extensively in all my systems. It's also a vital component of the Beneish M-Score (it's called TATA), a revised version of which I use in my universe rules to screen out a lot of stocks.

Standalone performance tells you next to nothing about how a factor is going to work in a comprehensive system. I even use some factors in the opposite way that a standalone performance would tell you to use, and the accrual ratio is one of them. What's most important about deciding whether to use a factor is understanding how it works from an accounting standpoint rather than looking at a standalone performance test.

I've put together a 10-factor ranking system with accruals as one of the ten nodes. (It's here: https://www.portfolio123.com/app/ranking-system/521620). The performance over the last 10 years on the easy-to-trade US universe with all nodes equally weighted is as follows:

Now here's the performance with the accruals node given 0 weight:

You can see the performance is better if you include the accruals node. But the accruals node as a standalone node has godawful performance over the same period and universe:

[rwbattyaz]

When this blog thread surfaced last August, I was inspired to try developing a screen using the concept. I implemented in early November and it has been rocky since with it cycling from fully funded, probationary (half weight), to defunded. It does have limited overlap with my other screens.

There does seem to be a definite universe impact. My standard starting universe is a liquidity constrained NOOTC based one:

IsMLP = False
MktCap > 50
Price > 3 OR Ticker("DTC:USA")
MedianDailyTot(20) > 200000
StaleStmt = 0
Spread(0)/Close(0) < .01
FOrder("StockID" , #PREVIOUS)

Please note the minimum market cap of $50 million as it seems to be important.

The screen I developed is:

Period 01/02/2021 - 07/12/2025
Rebalance Days 7

RULES
Common Between(FRank("MktCap" , #PREVIOUS, #DESC, #EXCLNA), 15, 70) // 15 70
Common Between(FRank("(NetIncBXorTTM-OperCashFlTTM)/AstTotTTM" , #PREVIOUS, #ASC, #EXCLNA), 70, 100) // 70 100
Common Between(FRank("(NetIncBXorPTM-OperCashFlPTM)/AstTotPTM" , #PREVIOUS, #ASC, #EXCLNA), 25, 90) // 25 90
Common Setvar(@FCapEx, FRank("CapExGr%3Y-CapExGr%5Y",#PREVIOUS,#DESC,#InclNA))
Common Between(@FCapEx, 15, 100) // 20 98
Common Between(Rating("Core: Value"), 30, 100) // 25 100
Common SetVar(@ShortInterest,(FRank("SI%Float" , #PREVIOUS, #ASC, #ExclNA)))
Common Between(@ShortInterest, 0, 100) // 0 100
Common Rank <= 96.0 // 96.0

SETTINGS
Method LONG
No Stocks 7
Slippage% 0.2500
CarryCost% 0.0000
Long Weight% 100.0000
Short Weight% 0.0000
Rank Tolerance 0.8000

The Ranking System used is "Core: Growth".

Somewhere along the line I came upon a remark stating that this factor works less well with larger companies which has the largest 30% excluded. Thinking that the factor may be overwhelmed in smaller companies the bottom 15% is excluded.

Another rule to note is the exclusion of the very best from the Ranking System as it steps up performance.

Simply plugging the Easy To Trade US universe into the screen has a 273% period return which does imply a universe impact.

Hope this helps a little.

Rich

[SpacemanJones]

Hi ScifoSpace. It's been a long while since I've worked with my ranking systems, but I recall accruals being effective when mixed in with other factors and do recall it not looking great by itself. I looked into one of my systems and I see accruals mixed into factors related to earnings and projected earnings. To speculate, I'm guessing it helps/helped balance out companies reporting or projecting relatively good earnings and/or earnings trends based on aggressive or conservative accounting. But it has been several years since I've really modified the ranking systems or further researched the factors, so ymmv. (Also, I just do the old school ranking factors and not sure how to do the new AI stuff, so maybe I have blind spot there.) all the best,