Company research
Reinvest or pay out: what each cohort actually grew into
Sort the largest US companies by how much of revenue they put into research, and the next six years of growth line up almost perfectly with the sort. Sort them by how much of profit they paid out instead, and the line runs the other way — but only in revenue, not in profit. Both results come with a caveat this page refuses to bury.
Sorted on FY2015–FY2019 · measured to FY2025
Sorted on FY2015–FY2019, measured FY2019–FY2025
- Highest research fifth
- 19.6% a year
- median R&D 25.8% of revenue
- Lowest research fifth
- 5.1% a year
- median R&D 0.8% of revenue
- Lowest payout fifth
- 8.2% a year
- median payout 19% of profit
- Highest payout fifth
- 6.2% a year
- median payout 105% of profit
Survivorship warning: this universe is the largest US companies today, so companies that failed between FY2015 and FY2025 are absent from every bucket. Growth levels here are flattered; the ordering is the part worth reading.
| Quintile | R&D range | Median R&D | Revenue growth | Middle half | Profit growth | Shrank | Companies |
|---|---|---|---|---|---|---|---|
| Q1least research | 0.0%–1.8% | 0.8% | +5.1%a year | 2.5 to 8.5% | +7.9% | 4 | 33 |
| Q2 | 1.9%–5.8% | 3.9% | +4.2%a year | 2.6 to 8.1% | +8.8% | 3 | 33 |
| Q3 | 5.8%–12.7% | 8.9% | +6.9%a year | 3.9 to 13.3% | +5.6% | 3 | 33 |
| Q4 | 12.9%–19.0% | 15.7% | +11.4%a year | 4.6 to 17.7% | +15.3% | 3 | 33 |
| Q5most research | 19.0%–467.4% | 25.8% | +19.6%a year | 12.4 to 36.7% | +16.4% | 2 | 35 |
| Quintile | payout range | Median payout | Revenue growth | Middle half | Profit growth | Shrank | Companies |
|---|---|---|---|---|---|---|---|
| Q1lowest payout | 0.1%–24.6% | 18.7% | +8.2%a year | 5.3 to 11.7% | +8.9% | 1 | 50 |
| Q2 | 24.7%–35.0% | 29.0% | +7.3%a year | 4.2 to 10.0% | +7.8% | 4 | 50 |
| Q3 | 35.1%–51.5% | 43.6% | +6.1%a year | 3.1 to 8.4% | +5.8% | 1 | 50 |
| Q4 | 51.6%–70.9% | 60.9% | +5.0%a year | 3.1 to 7.4% | +7.1% | 6 | 50 |
| Q5highest payout | 72.9%–390.5% | 104.7% | +6.2%a year | 4.3 to 10.4% | +7.9% | 6 | 53 |
| Group | Median revenue growth | Shrank | Companies |
|---|---|---|---|
| Both research and a dividend | +5.4% a year | 12 | 108 |
| Research, no dividend | +16.4% a year | 3 | 59 |
| Dividend, no research | +6.7% a year | 9 | 165 |
| Neither | +11.9% a year | 1 | 19 |
351 of 375 companies had enough history at both ends of the window to be sorted. Largest US-listed operating companies by market capitalisation, one line per company, foreign private issuers (ADRs and 20-F filers) excluded. Full provenance, method and a citation line are in Sources and method below. The universe and every filing behind them are listed in full.
Read the caveat before the table
The universe here is the largest US companies **today**. Every company that spent heavily on research, failed and disappeared between FY2015 and now is missing from it, and so is every company that paid out generously and shrank into irrelevance. That inflates the level of growth in all ten buckets on this page. It distorts the *comparison* between buckets only if the failure rate differed systematically across them — which it plausibly did, in the direction that flatters the research-heavy cohort. Treat the ordering as informative and the absolute numbers as generous.
With that on the table: sorting on FY2015–FY2019 research intensity produces a monotone ordering of the following 6 years of revenue growth. The lowest fifth compounded revenue at 5.1% a year and the highest fifth at 19.6%, and the gap survives inside the middle half of each bucket rather than resting on a few outliers. Monotone across all five buckets is a stronger result than a top-versus-bottom comparison, which is why the page shows all five.
The payout sort runs the other way and only halfway. Revenue growth falls from 8.2% in the lowest payout quintile to 6.2% in the highest — but profit growth does not follow the same path. High-payout companies kept compounding earnings through margins, cost control and a shrinking share count while their sales went nowhere. "Growth" turns out to be two different questions, and a dividend policy answers the first one better than the second.
What none of this establishes is causation. Research-heavy companies are concentrated in industries that were growing anyway, so part of the first sort is a sort on industry; high-payout companies are concentrated in mature industries, so part of the second is the same thing in reverse. The honest reading is that a company's spending pattern is a good description of where it already sits in its life cycle, and a much weaker claim about what a given dollar of research would do inside a different company.
Questions people ask about this
- Does this prove that spending on R&D causes growth?
- No, and the page is built so that the gap between the two claims stays visible. Research-heavy companies are concentrated in industries that were growing anyway, so the sort is partly a sort on industry. What the table establishes is that the association is large, monotone across all five buckets, and survives inside the interquartile range — which is more than most versions of this claim can say, and still not causation.
- What is the survivorship problem here?
- The universe is the largest US companies today. Every company that spent heavily, failed and disappeared is missing from the sample, and so is every company that paid out generously and shrank into irrelevance. That biases the level of growth in every bucket upwards. It biases the comparison between buckets only if failure rates differed systematically between them, which they plausibly did — so treat the ordering as informative and the absolute figures as flattering.
- Why does high payout look bad for revenue growth but not for profit growth?
- Because a payout ratio is profit in the denominator, and mature companies with steady profits and little revenue growth sit naturally at the top of that sort. Their profits kept compounding through buybacks, cost control and margin gains while their sales did not. It is a useful reminder that 'growth' is two different questions, and that a dividend policy tells you more about where a company already is than about where it is going.
Sources and method
- Data
- SEC filings (EDGAR) — Each company's annual report on Form 10-K. Every company named on the page links to the filing its figures were read from.
- Financial Modeling Prep — Distributor of the filing data, and the source of the market caps used to pick the universe.
- How it was calculated
- Companies are sorted into quintiles on their 2015-2019 average research intensity (R&D over revenue) and, separately, on their 2015-2019 average payout ratio (common dividends over net income, dividend payers only). Each quintile is then measured on realised compound growth in revenue and in net income from fiscal 2019 to the latest fiscal year — a forward window that does not overlap the sorting window. Medians rather than means, with the 25th and 75th percentile of revenue growth shown so the spread inside a bucket is visible. Profit growth is only computed where both endpoints are positive. The universe is the largest US-listed companies today, which means companies that failed during the period are absent; that limitation is stated on the page and is not removable from this data.
- How often it changes
- Rebuilt from the filings about once a year, after the bulk of the universe has reported. The fiscal year it runs through is at the top of the page.
- Citing this page
Free to quote — please link rather than copy the table.
Plutux. "Reinvest or pay out: what each cohort actually grew into." Data through FY2025. https://plutux.ai/ko/resources/tools/research-spending-vs-dividends
Historical figures for information only — not investment advice, and not a forecast.
Related tools
- R&D and capex intensityResearch spending and capital expenditure as a share of revenue, by sector, over a decade — with the denominator problem nobody else handles.
- Buybacks vs dilutionTen years of share issuance and repurchase, netted off — the part of a buyback that only cancels out the shares just granted.
- S&P 500 annual returnsEvery calendar year's price return, the largest drop inside each of those years, and what the distribution of outcomes actually looks like.