Key takeaways
- Across the 108 annual observations in its UK series, 1900 to 2015, the Jorda-Knoll-Kuvshinov-Schularick-Taylor study splits housing's 9.38% nominal total return into 5.44% of price growth and 3.94% of net rent.
- On the average UK house price of £273,000 in July 2026, stamp duty on an additional property comes to £17,300 against £3,650 on a sole home, a difference of £13,650.
- Councils in England can charge up to 2 times normal council tax on a second home. At the 2026-27 England average Band D bill of £2,392, that premium alone costs 0.88% of a £273,000 property every year.
- Six weeks of personal use is 0.37 percentage points of value a year at the 3.2% net rent-price ratio MSCI reported for UK residential property in 2013, worth £1,008 against a £2,392 premium.
- Private residence relief covers one home. On a £100,000 gain on the other, the 24% rate in force from 6 April 2026 takes £23,280 after the £3,000 allowance.
The second home total return is the housing return minus the rent you never collect
You've seen the research that puts long-run housing returns level with equities, and the obvious next thought is to buy another house. So what does the second one actually earn?
Less than the first, and the reason is arithmetic rather than opinion. The published housing returns are total returns: price growth plus the rent the building throws off. A second home you keep for yourself collects the price leg and almost none of the rent leg. Then it pays a stamp duty surcharge on the way in, a council tax premium every year it sits there, and capital gains tax on the way out, none of which apply to the home you live in.
Put the UK numbers through that filter and the second home total return lands near 4.4% a year nominal, against 11.25% on UK equities over the same historical sample. The gap isn't caused by houses being bad assets. It's caused by the owner consuming the part of the return that does the compounding.
Where housing's return came from: 5.44% in price and 3.94% in rent
The best long-run evidence is The Rate of Return on Everything, 1870-2015, by Oscar Jorda, Katharina Knoll, Dmitry Kuvshinov, Moritz Schularick and Alan Taylor, in the December 2017 San Francisco Fed working paper version. It covers bills, bonds, equities and housing across 16 advanced economies.
Their country table breaks each housing return into two parts. For the UK, whose housing series runs from 1900 to 2015, the arithmetic mean annual nominal figures are 5.44% of capital gain and 3.94% of rental income, a total of 9.38%, with 58.01% of the return coming from price. UK equities returned 11.25% nominal over the same table.
The United States tilts further. US housing, sampled from 1891, shows 3.54% of capital gain against 5.33% of rental income, a total of 8.87%, with only 39.94% of the return coming from price. US equities returned 11.08%. The chart plots all six of those figures.
One detail matters for what follows. Those rental yields are already net of the boiler. The authors state that their benchmark yields "reflect net income" that is "net of property management costs, ground rent, and other irrecoverable expenditure", and their cost chart carries the note "Costs include maintenance, depreciation, and other running expenses such as insurance. Taxes are excluded." So maintenance is inside the 3.94%. Council tax isn't.
A house price index measures one leg, and the UK HPI documentation says so
The house price indices a UK reader sees quoted are price indices. The official one "captures changes in the value of residential properties from a base of 100 set in January 2015," using hedonic regression to hold the mix of property types constant. It's a careful measure of what a comparable house sells for. It says nothing at all about rent.
That's not a flaw in the index. It's a mismatch between what gets published and what people compare it to. A house price index measures what a building sells for, never what it earns, because it was not built to price an investment. Reading a 5.44% price series as though it were the whole return is the same error as reading a share price chart and forgetting the income, except that with housing roughly 42% of the return is the part you dropped.
Entry: £17,300 of stamp duty on an average home, against £3,650
Take the average UK house price for July 2026, which the Office for National Statistics put at £273,000. Buy that as your only home in England and stamp duty land tax runs at zero up to £125,000, 2% on the portion to £250,000 and 5% above it. The bill is £3,650.
Buy the same house when you already own another and every band moves up by 5 percentage points: 5%, 7% and 10% across those same slices. The bill is £17,300, or 6.34% of the purchase price. The surcharge is £13,650, and because it applies to the whole price rather than the top slice, it's exactly 5% of what you paid.
Second home stamp duty behaves differently from an annual cost. Spread over a ten-year hold, that one payment is worth half a percentage point a year off the return. Over five years it's a full point. It's also cash that never enters the asset, which is a different thing from a cost that reduces the asset's yield. The same distinction drives the mortgage overpayment vs investing comparison, where what matters is the rate a pound earns net of the tax it has already suffered.
Holding: a council tax premium of up to 100%, on top of the costs the yield already nets off
From 1 April 2025, councils in England gained a new power. GOV.UK puts it plainly: "You can also be charged up to 2 times your normal Council Tax. This is sometimes known as the second homes premium." A furnished property with nobody living in it as their main home qualifies.
The average Band D bill set by English local authorities for 2026-27 is £2,392, including adult social care and parish precepts. Double that and the premium adds £2,392 a year. On a £273,000 property that's 0.88% of value, every year, before a single repair.
It isn't a theoretical power either. The council taxbase count for 10 September 2025 recorded 268,000 dwellings in England classed as second homes, out of 25.8 million dwellings in total. Of those, 170,000 were charged the newly introduced second homes premium. Most English second homes are paying it.
Remember what the Jorda net yield already contains and what it doesn't. The holiday home costs people worry about most, maintenance, depreciation and insurance, are inside the 3.94%. Taxes are outside it. The premium is an addition to the cost side, not a restatement of it.
Exit: private residence relief covers one home, and 24% applies to the other
HMRC's position is short. You may owe capital gains tax "when you sell (or 'dispose of') property that's not your home." The capital gains rates guidance adds, separately, that there isn't usually tax on the disposal of your own home. One property gets the relief. The rest don't.
From 6 April 2026 a higher or additional rate taxpayer pays 24% on gains, and a basic rate taxpayer pays 18% within the basic band and 24% above it. The annual exempt amount for 2026-27 is £3,000. On a £100,000 gain that leaves £97,000 chargeable and £23,280 payable at the higher rate.
The study's own property tax appendix records the same point for the UK, writing that "No capital gains tax is payable if the property was the owners' principal residence." Every long-run housing return you've read is measured on a national housing stock in which the typical unit is somebody's only home. Your second one isn't.
What weeks of use are actually worth
Here's the part people skip. If you use the place yourself, you do collect part of the rent leg, just not in cash. You collect it as accommodation you'd otherwise have paid for. The question is how much.
The Jorda appendix gives a modern benchmark: for 2013, MSCI reported a rent-price ratio for UK residential real estate of 0.032, net of running costs. Apply 3.2% to a £273,000 property and the building throws off £8,736 of shelter a year, which is £168 a week across all 52 weeks.
Use it six weeks a year and you've captured £1,008, or 0.37 percentage points of the property's value. The council tax premium alone is £2,392, so the shortfall on that single line is £1,384. At six weeks of use and a full premium, the income leg of a second home isn't small. It's negative, at roughly 0.51 percentage points a year.
To capture the whole £8,736 in six weeks, those weeks would need to be worth £1,456 each. That is the honest break-even, and it is not obviously out of reach for a coastal cottage in August. It is well out of reach for six weeks in February. For scale, the average monthly private rent in the UK was £1,400 in August 2026, which is about £323 a week for a whole year of occupancy rather than a peak fortnight.
Assemble it. The 5.44% price leg, minus 0.51 points where the rent leg used to be, is 4.93%. Amortise the £13,650 surcharge over ten years and it's 4.43%, before any capital gains tax on the eventual sale. UK equities in the same table returned 11.25%. Both figures are nominal, unlevered and historical, and neither is a forecast.
The strongest case against this arithmetic
There's a serious objection, and it comes from the same paper. Risk-adjusted, housing wins, and it isn't close. The authors write that "Housing provides a higher return per unit of risk in each of the 16 countries in our sample, with Sharpe ratios on average more than double those of equities." The same paper puts the standard deviation of global real housing returns at 9.98% against 21.94% for equities. Those two are real rather than nominal, so they do not sit on the same scale as the return figures above, but the ordering is the point. A 4.43% return at housing volatility is not obviously worse than 11.25% at equity volatility, and a national index understates how differently one building behaves from the average.
Second, the price leg itself may be understated for the properties people actually buy as second homes. Nationwide's analysis found that a property inside a National Park attracts a 24% premium over a similar property elsewhere, around £66,500 against the Q2 2026 UK average price of £278,784, with a 6% premium for anything within 5km of one. Nationwide points to controlled development and limited new construction as part of the reason, which is a different mechanism from the one driving a national average.
Third, the usage value is real. £1,008 of accommodation you didn't buy is £1,008, and it is the one part of the return that arrives without a transaction. A family that genuinely spends eight or twelve weeks a year there is collecting £1,344 or £2,016 on the same 3.2% basis, which changes the sign of the income leg.
Fourth, you can let it when you're not there, which converts usage value back into cash rent. That route got more expensive. HMRC confirms the furnished holiday lettings rules "cease to apply in tax years commencing on or after 6 April 2025," removing the finance cost exemption, the capital allowances treatment, the chargeable gains reliefs and the counting of that income as relevant earnings for pension relief. A let second home is now taxed like any other rental, which is the territory covered by buy-to-let Section 24.
What this arithmetic cannot tell you
Start with the vintages, because they don't match. The 5.44% price leg is a 1900 to 2015 UK average that contains decades of high inflation, and it's nominal. The 3.2% rent-price ratio is a 2013 level applied to a 2026 price. Stitching a historical return series to current tax rates produces an illustration, not a projection, and a different century would produce a different number.
The housing series is a national index of the whole stock, unlevered, with no mortgage in it. Most second homes are bought with debt, which magnifies the price leg in both directions and adds an interest cost this model doesn't carry. The equity figure is a national index too, before dealing costs and before tax.
The council tax second homes premium is a power, not a rate. Councils set it at their discretion between nothing and 100%, and 98,000 of England's 268,000 second homes weren't charged it at all in the September 2025 count. Scotland's rules differ, and GOV.UK lists exceptions for annexes, job-related accommodation, planning-restricted properties and homes being marketed for sale.
The usage-value calculation also assumes weeks are interchangeable, which they clearly aren't. A rent-price ratio is an average over a year of continuous occupancy. Nobody buys a holiday home to occupy it in the average week. The £1,456 break-even is the right question to ask, but the answer is specific to a building, a location and a month, and no national dataset can give it to you.
What would change the conclusion
If the surcharges went, most of the gap closes. The 5% stamp duty surcharge and the doubled council tax are policy, not physics, and the council tax premium only became available to English councils on 1 April 2025. Remove them and the second home total return moves from about 4.43% back toward the 5.44% price leg, which is a different argument entirely.
If you let it out properly, the model stops applying, because the rent leg comes back as cash and the comparison becomes a rental yield calculation rather than a usage one. That's a business with tenants, voids and the post-2025 tax treatment attached, and it's closer to the REITs vs direct property question than to this one.
If the price leg diverges from the national index, the whole calculation shifts. The 24% National Park premium is evidence that constrained locations can behave differently over long periods, and the second-home market is concentrated in exactly those places. A national series cannot price a specific coastline.
What's worth watching isn't the house price index. It's the two lines that never appear in it: what the weeks you actually use are worth to you, and what your council decides to charge for the ones you don't. Those two numbers, not the index, decide what the second property returns. The same gap between a published index and a held asset runs through your house as an investment, where the rent you consume as shelter never shows up as a return either.