The question
Most comparisons of this kind measure returns. The question that comes first, and turns out to matter more, is what £20,000 could actually reach.
Suppose you had £20,000 in January 2015. Three ordinary ways of trying to make it grow would have been the stock market, a rental property, or a business.
The window runs to January 2026, eleven years, because that is the only period over which every leg can be measured rather than estimated. Method and assumptions are in Appendix A, and the corrections made along the way are in Appendix B.
Door one: the index fund
The door was fully open, and the only decision that separated the two outcomes was made once, at the start.
Two real accumulation funds rather than indices, because you cannot buy an index. Accumulation share classes mean dividends are reinvested inside the fund price rather than paid out, so every figure below is a total return.
| Fund | Jan 2015 | Jan 2026 | Total | Annualised | £20,000 becomes |
|---|---|---|---|---|---|
| CT FTSE All-Share Tracker 2 Acc | 5.329 | 11.65 | 118.6% | 7.4% | £43,723 |
| Fidelity Index World P Acc | 1.1532 | 4.1582 | 260.6% | 12.4% | £72,116 |
Sources: published fund prices. FMF Analysis calculations.
- Accumulation share classes, so dividends are reinvested inside the price. Net asset value, before platform fees.
Both funds are passive. Both required the investor to do nothing except not sell. One returned two and a half times what the other did. That is the first thing worth noticing, and it recurs in every section that follows.
A fund has no minimum purchase worth worrying about, no deposit and no lender to satisfy: whatever you have is what you can invest. What it demanded was the ability to watch a fall without selling. The FTSE All-Share's worst drawdown over the past decade was 35.3%, most of it inside a few weeks in early 2020.
For context, over the ten years to mid-2026 the FTSE All-Share returned about 6.1% a year against 11.9% for the FTSE All World. The 7.4% above is higher because this window captures the strong UK year in 2025, so the comparison that follows is mildly generous to UK equities.
Door two: the rental property
This door was shut in the market people mean when they say buy-to-let, and within reach almost everywhere else. The size of the miss varied by a factor of fifty.
What £20,000 could reach
At 75% loan to value, using quoted rates published by the Bank of England, the cash needed to buy the average flat in January 2015 was:
| Flat | Deposit | Stamp duty | Legal | Cash required | Short of £20,000 by |
|---|---|---|---|---|---|
| North East | £19,581 | none | £2,000 | £21,581 | £1,581 |
| North West | £22,434 | none | £2,000 | £24,434 | £4,434 |
| Yorkshire and The Humber | £22,640 | none | £2,000 | £24,640 | £4,640 |
| Wales | £22,732 | none | £2,000 | £24,732 | £4,732 |
| England | £42,596 | £908 | £2,000 | £45,504 | £25,504 |
| London | £93,803 | £8,761 | £2,000 | £104,564 | £84,564 |
Sources: HM Land Registry, Bank of England. FMF Analysis calculations.
- Deposit at 25% of the purchase price, at 75% loan to value.
Scroll the chart sideways to see it in full.
Sources: HM Land Registry, Bank of England and FMF Analysis calculations.
- Deposit at 25% of the purchase price, plus stamp duty land tax at the rates then in force and £2,000 of legal and survey costs.
- No stamp duty was due on the four cheapest regions, all of which sat below the £125,000 threshold in force at the time.
In the North East the shortfall was £1,581. In London it was £84,564. No stamp duty was due on the four cheapest, because all four sat below the £125,000 threshold in force at the time.
The choice inside the choice
Over these eleven years, flats were the worst-performing property type in every single UK region. Not most. All thirteen, plus every national aggregate.
| Region | Detached | Semi | Terraced | Flat |
|---|---|---|---|---|
| Northern Ireland | 86.8% | 88.4% | 96.4% | 69.0% |
| North West | 63.8% | 73.8% | 75.2% | 54.1% |
| West Midlands | 67.9% | 74.7% | 72.9% | 47.8% |
| England | 52.6% | 60.1% | 58.3% | 28.1% |
| London | 40.3% | 46.4% | 43.6% | 14.9% |
Source: HM Land Registry, average price by property type.
- Five of thirteen regions shown. Flats were the weakest type in all thirteen, plus every national aggregate.
London was also the worst-performing region for every property type, not just flats. A London flat was therefore the single weakest cell in the entire grid: worst type, worst region. Which matters, because a London flat is exactly what most people picture when they say buy-to-let.
What it returned
A note on the figures that follow. The prices, rents and mortgage rates are official published data. Everything built on top of them is a hypothetical scenario using the best data available to me, with the assumptions listed in Appendix A. Better data may exist, and different assumptions would produce different figures.
Take the London flat first, since it is the version of this trade that gets discussed. The flat cost £375,213 in 2015 and was worth £431,082 by January 2026, a rise of 14.9%. The £281,410 loan is interest-only, which is the standard structure rather than a convenient assumption: the Bank of England reports 82% of buy-to-let loans are interest-only, against 11% of owner-occupier loans.
Scroll the chart sideways to see it in full.
Source: Bank of England series IUMZID4.
- The model refinances every two years in January, five times, followed by a one-year stub to reach the January 2026 exit.
- Quoted rather than effective rates. The rate actually paid on an existing loan will differ from the rate advertised on a new one.
Refinancing every two years at the Bank's quoted rate produces £103,474 of interest across eleven years. Rent for a London flat ran from £1,390 a month to £1,987, collecting £210,706 gross over 132 months. After letting fees, voids, maintenance and that interest, £27,493 of net rental cash remains. Selling and repaying the loan leaves £141,706 of equity.
Total £169,199, or 1.62 times the £104,564 invested. The same capital in the UK tracker returned 2.19x and in the global tracker 3.61x. On this evidence alone, property loses.
Running the same model elsewhere
Identical assumptions, regional prices, regional rents by property type, the same financing and the same costs. The answer changes completely.
Section 24, phased in from April 2017 and fully in force from April 2020, stopped landlords deducting mortgage interest before tax and gave them a basic rate credit instead, which costs a higher-rate taxpayer considerably more.
| Property | Cash required | Growth | Entry yield | Multiple | After Section 24 |
|---|---|---|---|---|---|
| North West terraced | £26,608 | 75.2% | 7.2% | 4.94x | 4.18x |
| England terraced | £41,009 | 58.3% | 7.2% | 4.40x | 3.64x |
| Wales terraced | £26,995 | 69.0% | 6.4% | 4.33x | 3.73x |
| North West flat | £24,434 | 54.1% | 6.8% | 3.95x | 3.27x |
| North East flat | £21,581 | 28.2% | 6.9% | 2.95x | 2.28x |
| London flat | £104,564 | 14.9% | 4.4% | 1.62x | 1.32x |
| Global tracker | any | n/a | n/a | 3.61x | 3.61x in an ISA |
| UK tracker | any | n/a | n/a | 2.19x | 2.19x in an ISA |
Sources: HM Land Registry, ONS, Bank of England. FMF Analysis calculations.
- The Section 24 column assumes a taxpayer on the higher rate. Funds held in an ISA are unaffected.
- Growth is the change in the average price for that type and region over the window, not the return on the cash invested.
Scroll the chart sideways to see it in full.
Sources: HM Land Registry, ONS, Bank of England and FMF Analysis calculations.
- Large marker before personal tax, small faded marker after Section 24 for a taxpayer on the higher rate. The funds are shown inside an ISA, so no adjustment applies to them.
- The two funds are a sample and not the range of what equities returned. Other indices over the same eleven years returned considerably more, and others considerably less.
- Property assumes 75% loan to value, interest only, refinanced every two years, letting agent at 10% of rent, voids at 5%, maintenance and insurance at 1% of value a year, and rent that is not reinvested. Those costs are roughly double the Bank of England's benchmark, so the property figures sit at the pessimistic end of a plausible range.
Every property in that table beat the UK tracker. Three of them beat the global tracker even after Section 24. The London flat lost to everything, including the version of itself bought 200 miles north for a quarter of the money.
Two things drove the gap, and neither was sophisticated. Northern property appreciated three to five times as fast. And it yielded around 7% at purchase against London's 4.4%, so the rent covered the borrowing comfortably rather than barely.
The yield gap is not a quirk of the sample. The Bank of England puts average buy-to-let gross yields at around 6% since 2014. The northern figures sit above that. London sits well below it. A London landlord accepted a materially worse income return, for the same borrowing cost, in exchange for capital growth that then failed to arrive.
The lender's own test
There is a better way to judge whether these purchases were sound, and it is not this model's arithmetic.
Buy-to-let affordability is assessed on the interest coverage ratio: rent divided by mortgage interest at a stressed rate. Under the Prudential Regulation Authority's supervisory statement SS13/16, lenders stress at the higher of 5.5% or two points above the mortgage rate, and require coverage of at least 125%, rising to 145% for higher-rate taxpayers. These rules are not history. SS13/16 took effect in January 2017 and remains the framework lenders work to today, most recently amended with effect from January 2026. The thresholds have not moved.
| Property | Annual rent | Interest at 5.5% | Coverage |
|---|---|---|---|
| North West terraced | £7,116 | £4,060 | 175% |
| North East flat | £5,436 | £3,231 | 168% |
| North West flat | £6,072 | £3,702 | 164% |
| London terraced | £23,028 | £18,455 | 125% |
| London flat | £16,680 | £15,478 | 108% |
Sources: ONS, HM Land Registry, Prudential Regulation Authority.
- The PRA requires coverage of at least 125%, rising to 145% for higher-rate taxpayers, stressed at the higher of 5.5% or two points above the mortgage rate.
The London flat comes out below the 125% minimum. The London terraced house lands exactly on the line. Everything further north clears it with room to spare.
Our £20,000 buyer could never have bought either London property, so why does this matter to them? Because it explains what they were actually being kept out of. The northern flat they could nearly afford was a fundamentally sounder piece of business than the London one they could not: it generated enough rent to survive a rate rise, and the London flat did not. Being priced out of London was not being priced out of the good version of this trade.
The ratio measures something real regardless of when it was codified: whether rent covers borrowing with enough slack to absorb higher rates. The London flat had none from the day it was bought, and 2022 was precisely the event it had no slack for. The regulator's test and the eleven-year return arrive at the same answer independently.
Door three: the business
This door was wide open, and it is the one about which almost nothing can be said with a number.
£20,000 is not a deposit on a business; it is capital. It starts a trade, buys a van and a customer list, or takes a stake in something already running. In the London property market it was a rounding error. Here it is a real sum.
There is no price series for a small business, because businesses are not fungible. A café, a plumbing firm, a laundrette and a software company are not units of the same asset. What the official data can tell us is how often businesses stayed open.
Of businesses born in London in 2015: 86.4% survived one year, 68.2% two, 51.8% three, 42.8% four and 36.7% five. Across the 2010 to 2018 cohorts, five-year survival in London ranged from 36.7% to 41.7%, averaging 39.1%. The North West averaged 38.9%.
Scroll the chart sideways to see it in full.
Source: ONS Business Demography via the London Datastore.
- Shaded band shows the highest and lowest five-year survival across the 2010 to 2018 cohorts, for London and the North West.
- Closure is not failure, and survival is not a return. The ONS defines an active business through employment and turnover, and records neither profit nor the price a business sold for.
Using nine cohorts matters, because the 2015 cohort's fifth year was 2020. The neighbouring cohorts show the picture does not depend on the pandemic.
So roughly three in five closed within five years. That statistic is usually where these discussions end, and it is where they go wrong.
Closure is not failure. The ONS recorded around 280,000 business deaths in 2024. The Insolvency Service recorded 23,938 company insolvencies in England and Wales in 2025, a rate of one in 189 companies, and explicitly excludes solvent closures from its count. The two numbers differ by roughly a factor of ten. Most businesses that close are not collapsing. They are being retired, sold, merged, or simply stopped.
And the population was growing throughout. In London, business births exceeded deaths in every year from 2015 to 2022. The active business stock rose from 541,310 in 2015 to 592,330 in 2023, an increase of 9.4%. The North West grew 9.8% over the same period. So the honest description is high churn inside a growing population, not a graveyard. Individual businesses turned over rapidly; the number of them went up.
What none of this tells us is what an owner earned. The ONS defines an active business through employment and turnover. It does not record returns. A business leaving the statistics is not an investor losing everything, and a surviving business is not necessarily a good investment. There is no honest way to convert survival into a return, so none is attempted here.
That leaves the third door in an odd position: the one most open to £20,000 is the one about which the least can be said.
What each door demanded
Return is only one of the things these three doors differed on, and not the one that decided which were available.
| Index fund | Rental property | Business | |
|---|---|---|---|
| Minimum capital | none | £21,581 to £104,564 | £20,000 is real capital |
| Time required | none | ongoing | usually all of it |
| Liquidity | daily | months to sell | very low |
| Cash while held | none unless sold | monthly, if tenanted | variable |
| Exposure to rule changes | low | high | moderate |
| Can the return be measured? | precisely | precisely | no |
Source: FMF Analysis.
- Rule changes affecting rental property over this period included Section 24, the stamp duty surcharge and selective licensing.
Conclusion
The starting hypothesis was that reasonably risky assets converge towards similar outcomes given enough time. This period does not support it, but it fails in a more specific way than a simple ranking would suggest.
The spread inside each category was wider than the spread between categories. Property ranged from 1.62x to 4.94x depending on where and what you bought. Equities ranged from 2.19x to 3.61x depending on whether you looked at Britain or the world, and those two funds are only a sample: other indices over the same eleven years returned considerably more, and others considerably less. In both cases, the decision people agonise over, shares or bricks, mattered less than a decision they often make without thinking: which shares, which street.
The London flat was not property. It was one cell in a grid of sixty. Worst-performing type, in the worst-performing region, at the lowest yield, with the tightest interest coverage. It has become the default mental image of a UK buy-to-let, and over these eleven years it was the worst version of the trade available.
And the doors that could be measured were not the doors that were open. £20,000 comfortably bought a fund and could genuinely start a business. It did not buy a rental property anywhere in the country, though in the North East it came within £1,581. The fund and the flat can be tracked to the pound eleven years later. The business, the only one of the three where £20,000 was a serious sum rather than a fraction of one, cannot be tracked at all.
Which is worth remembering when reading comparisons like this one, including the parts of this one that came easily. The assets that generate clean data get compared endlessly. That is not the same as their being the best available, and it is not the same as their being within reach.
None of this is a recommendation, and no winner is declared. It describes one eleven-year window using observed data and a stated set of assumptions, chosen before the results were known and listed in full below. A different window, or different assumptions, would produce different figures.