There is a version of value creation that is really value extraction wearing a suit. Optimise the operation. Strip the fixed assets. Push the cost line down, quarter after quarter. Revenue holds, margin improves, the share price responds, and everyone involved is rewarded for it.
It works. That is the difficult part. It works for long enough that by the time it stops working, the people who did it have been promoted, and the people left holding the company cannot say precisely what was taken.
What was taken was time.
New products do not come from a line item. They come from people who have enough room to look at a problem for longer than the problem strictly requires, and who are not so afraid of next quarter that they cannot afford to be wrong on the way to being right.
The head of Bell Laboratories, writing on innovation in the 1930s, described the conditions rather than the process: an environment where people did not fear for their jobs, and could focus and learn. That was not sentiment. It was an operating specification for the most productive industrial research lab of the century.
Every efficiency programme is, among other things, a slack-removal programme. That is what makes it effective, and it is also what makes it expensive in a currency nobody records.
Take out the redundancy and the organisation gets cheaper and more brittle at the same rate. The cost shows up years later, as an inability to respond, and by then it is attributed to market conditions.
You do not have to imagine this. An industry has run the experiment at scale and published the results.
Read those as three separate operational facts and they are a familiar complaint about legacy systems. Read them together and they describe something more specific: an industry that optimised itself into immobility. Seven-tenths of the capacity is committed before anyone proposes anything. Half of what is proposed gets cancelled. What survives takes a financial quarter to ship.
No individual decision in that sequence was wrong. Each was a defensible efficiency. The aggregate is an organisation that can no longer do the thing it exists to do.
You cannot cut your way to a new product. You can only cut your way to being unable to make one.
A cost reduction has a line. A foregone product does not. There is no ledger entry for the thing that was never built, no variance report for the market entered two years late, no write-down for the capability quietly traded away in a restructuring that came in under budget.
So the measurement system rewards exactly one half of the transaction. The savings are booked; the cost is unbooked and arrives later, attributed to something else. This is not a failure of will. It is a failure of measurement — and it is the reason the trade looks free at the moment it is made.
The usual framing treats productivity gains as the modest prize and revenue growth as the real one. The causal order runs the other way.
Recovered capacity is what produces faster time to market. Faster time to market is what produces new markets, extended product lifecycles, and the earnings that fund whatever comes next. Every revenue outcome worth having sits downstream of somebody having the time to build it.
Which is why a business case promising revenue directly gets discounted by any competent CFO. The mechanism is missing. A business case promising recovered capacity, measured against a baseline, is checkable — and the revenue arrives as its consequence.
It also explains why extraction is so hard to reverse. You are not restoring a cost line. You are rebuilding a capability, with people who have learned that proposing something is how you get blamed.
Run that incentive across a market and you get a pattern worth sitting with: earnings expand while the capacity underneath them contracts, and the index records only the first half.
Capital returned through buybacks lands in earnings per share this year. Capital committed to research lands, if it lands at all, in a product several years out, under a different management team, measured by nobody currently being compensated. Faced with that asymmetry, a rational executive with a three-to-five year tenure does not have a hard decision to make.
Whether aggregate innovation is genuinely shrinking is a live argument among economists, and the measures disagree with each other. The incentive structure is not in dispute. Extraction is booked quarterly; the foregone product has no line, no variance report, and no owner.
The economist William Lazonick has spent a career on it, and his vocabulary is almost word for word the distinction above. He calls the two regimes retain-and-reinvest — earnings held back and put into productive capability — and downsize-and-distribute, where they go to financial interests instead. He calls the first value creation and the second value extraction.
His accounting of what the S&P 500 actually did with its earnings is the part worth sitting with:
Fifty-five and forty-one leaves four. Four percent of a decade of S&P 500 net income was available for productive capability, research and employee income combined. And the direction of travel is the point: across the previous decade, 2003–2012, the same accounting came to 91% — 54% to buybacks and 37% to dividends across 449 companies. Ten years on it is 96%, and the buyback total has gone from $2.4 trillion to $5.7 trillion.
In the early 1980s, before buybacks became routine, the figure was low single digits.
Sector by sector it gets sharper still. Among the fourteen pharmaceutical companies in that same dataset, distributions to shareholders reached 110% of net income — more than the companies earned, funded from reserves or borrowing. That is an industry whose entire premise is a research pipeline.
Lazonick’s claim is that innovation is not a budget line but collective and cumulative learning inside an organisation — which needs retained earnings and retained people, and is therefore precisely what a distribute-first regime spends.
A second strand of the literature, on declining business dynamism, tracks the other end: fewer new firms, fewer high-growth young firms, less reallocation of people toward productive uses. Different data, adjacent conclusion.
The argument is not settled — there is a serious counter-case that buybacks return capital so it can be reallocated by the market, and that research spending has risen in absolute terms. Read both. But if you find yourself arguing that extraction has no cost, notice that you are making a claim about a number nobody measures.
Which produces the uncomfortable possibility that an index can rise for years on the strength of companies steadily becoming less able to make new things — and that the rise is not evidence against the problem, but the mechanism of it.
It is the same unbooked trade described above, made at the level where nobody has to watch it happen. At a single company, somebody eventually notices that the product pipeline is empty. At index level, the company that stopped innovating is simply replaced by one that has not yet started.
An astute reader stops here and names the counterexample: the largest buyback programmes in history belong to Apple, Microsoft, Alphabet, Meta and Amazon — companies nobody would accuse of operational immobility. If distributing 96% of earnings causes an inability to innovate, the top of the index should be the least innovative part of it. It plainly is not.
Two distinctions do part of the work. A great deal of mega-cap buyback volume is dilution offsetting rather than capital evacuation — neutralising the share issuance from stock-based compensation, so the count stays flat. That is a payroll cost routed through the equity line, and it is a different act from shrinking the share count to lift earnings per share while cutting research. And an index average conceals enormous dispersion: a handful of extraordinarily profitable companies can distribute freely and still fund everything, while the remaining four hundred and fifty cannot.
But the real answer is that the ratio was never the right unit. Apple’s retained 4% is a larger absolute sum than most companies’ entire revenue. A mature insurer’s retained 4%, on ordinary margins, is not enough to modernise a policy administration system.
Which sharpens the claim rather than weakening it. The thesis is not that a distribution ratio predicts immobility. It is that capacity consumed without being measured predicts immobility — and capacity is an absolute quantity, not a percentage. A firm with extraordinary margins has so much slack that even aggressive distribution leaves plenty. A firm with ordinary margins distributing at the same rate has nothing left to renew itself with, and no line item recording that it has run out.
So the narrow version, which is the one the evidence actually supports: for any company without exceptional margins, distributing ninety-six percent of earnings leaves insufficient absolute capacity to fund its own renewal. That is not a claim about the index. It is a claim about the several hundred companies in it that are not monopolies — the insurers, the manufacturers, the mature industrials — and it is exactly where the operational evidence above comes from.
A percentage without a baseline tells you nothing. That is true of a vendor’s ROI claim, and it is true of this one.
Take a baseline before you optimise. Not a budget — a measure of capacity. How many hours go to keeping things running versus changing them. How much of the analyst's week is coordination rather than solving. How long a change takes from decision to live.
Then hold the efficiency programme to both numbers. Cost went down: by how much, verified. Capacity went down: by how much, verified. If you only measure one side, you have not made a decision. You have made a trade with the half you cannot see.
An ROI lens does not ask you to choose between creating value and extracting it. It asks you to see both — so the trade becomes a decision somebody made, rather than a transfer nobody recorded.
Optimise both and the arithmetic stops being zero-sum. Shareholders get returns that compound instead of returns that consume. Employees keep the capability they built. Customers get the products that capability produces. None of that requires anyone to be generous — only for the second number to exist.
Get the ROI lens on your numbers →