Property earns its place through contracted income from a physical thing, driven by different forces from a share. But its low measured volatility is partly an artifact of being appraised rather than traded, and that artifact has a size you can compute.
9 min read · Level 2
There is a version of the case for property that I think is quietly false and extremely popular, which is that it is a calmer asset than equities. It is not calmer; it is quieter. The events are the same size and they are reported less often, and a portfolio that feels steadier for holding it is frequently a portfolio that has simply arranged not to be told. That sounds like a criticism and I am not sure it entirely is, because the behavioural benefit is genuine and substantial: an owner who is not shown a daily price does not sell into a panic, and selling into panics is the single largest destroyer of individual returns there is. So the smoothing does real good, by accident, for reasons nobody would defend out loud. What it must not do is enter the analysis. The moment a smoothed volatility figure goes into an allocation decision alongside a marked one, the method is comparing a measurement with a measurement of a measurement, and it will favour the quieter one every time - not because the asset is better but because the camera was pointed at it less often. The defensible reason to hold property is that its income comes from different forces than everything else you own, and that reason is unaffected by how the asset is recorded. If that sentence cannot be written for a particular holding, no statistic should be allowed to make the case in its place.
The job property does that the rest of a portfolio does not comes down to three things, and only three.
Contracted income from a physical thing. A tenant owes a stated amount on stated dates under a lease. That is more certain than a dividend, which a board declares at its discretion, and less certain than a coupon, which is a debt obligation. Sitting between the two is a genuinely distinct position, and it comes attached to an object that exists whether or not anybody is currently willing to buy it.
A different set of drivers. What determines Property P's rent is local employment, the supply of competing units on that street, transport, and how many people want to live there. Those are not the forces that set a global technology company's earnings. A portfolio whose holdings all respond to the same drivers has fewer positions than it has line items, which is the whole of why-diversification-works, and property genuinely responds to something else.
Some relationship with the price level. Rents are renegotiated periodically and building costs rise with everything else, so over long periods this class tends to move with the general price level in a way a fixed coupon cannot. That is a tendency rather than a mechanism, and it has failed over meaningful periods, but it is a real reason to hold a real asset.
And the honest half, which is this lesson's actual contribution.
Property is not marked. It is appraised, infrequently, and each appraisal is anchored partly to the last one - which is exactly what lesson three's comparables method produces. So the recorded series is a smoothed version of whatever the true values were doing, and every risk statistic computed from that series inherits the smoothing.
The effect has a size. Take a true annual series of +14%, -9%, +6%, +21%, -17%, +8%, and record it the way an appraisal-based index does, with each year's recorded value half the current true value and half the previous recorded one:
True +14.00 -9.00 +6.00 +21.00 -17.00 +8.00 Recorded +7.00 -1.52 +2.18 +11.79 -4.12 +1.68
Standard deviation of the true series 14.27% Standard deviation of the recorded series 5.77% The recorded figure is 40.4% of the true one
Worst true year -17.00% Worst recorded year -4.12% which is 24.2% of it
Because that table changes what a great many published comparisons of this asset class actually mean.
Most statements about property's place in a portfolio rest on measured numbers: it has been less volatile than equities, it has had a low correlation with them, it has produced a good return for the risk taken. Every one of those is computed from a recorded series, and the recorded series in the table above showed 5.77% of volatility where the underlying reality had 14.27%. Nothing was falsified. The recording method simply cannot show a fall the appraisers have not yet seen evidence for.
Correlation is affected too, though less dramatically, and it is worth being precise rather than overstating it. Take a market series of +18%, -12%, +9%, +26%, -21%, +11% alongside the two property series above. The correlation between the market and the true property series is 0.999. Between the market and the recorded series it is 0.936. Both are high; the recorded one is lower.
So the artifact is real and it is not evenly distributed. Smoothing crushes measured volatility far harder than it crushes measured correlation - to 40.4% of the true figure against a drop from 0.999 to 0.936. The practical consequence is that the most misleading number in a property comparison is usually the risk figure rather than the correlation, which is the reverse of where most attention goes.
None of which means property is not diversifying. It means the case has to rest on something other than the statistics. The drivers genuinely are different: an office block's rent does not respond to a semiconductor cycle. That is a structural argument and it survives the smoothing problem entirely, because it was never a claim about a correlation coefficient in the first place.
What does not survive is the comfortable version, in which property is a low-volatility asset that quietly does well. It is a volatile asset that is measured rarely, and the two produce very different-looking charts from the same events.
There is one more thing that pulls this class back towards everything else at the worst moment. Property is bought with borrowed money, so it is exposed to credit conditions as well as to rents. When lending tightens, buyers who were relying on a mortgage disappear, and that happens at the same time as everything else in a portfolio is under pressure. Whatever independence the class has in ordinary periods is weakest in exactly the period diversification is supposed to help, which is the correlation-breakdown pattern correlation-between-assets describes.
Educational, not personalised financial advice. Nothing in the Investor Academy is a recommendation to buy, sell or hold anything — what is right for you depends on your own circumstances and the rules where you live.
Two portfolios, both described by their owners as diversified across asset classes.
Portfolio one. Listed shares, government bonds, and a listed property trust holding two hundred buildings. Everything is marked continuously. In a bad quarter the owner watches all three fall, sees the property holding fall too, and concludes property offered no protection.
Portfolio two. Listed shares, government bonds, and one directly owned building. In the same bad quarter the shares and bonds fall visibly and the building's recorded value does not move at all, because no comparable sale has completed and no appraisal is due. The owner concludes property did offer protection.
The buildings in both portfolios were subject to the same conditions. The difference in the two experiences is entirely a difference in reporting frequency, and the second owner's conclusion is the one that will be repeated at dinner.
This is not an argument for preferring the listed version. It has its own genuine drawbacks - the company's own debt and management decisions, a share price that can trade away from the underlying value, and the fact that visible falls cause holders to sell at the wrong moment more often than invisible ones do. The direct owner's insulation from a daily quote is a real behavioural advantage, and it is one of the honest arguments for the direct route.
The argument is only against using one version's measured statistics to justify holding either. If a portfolio holds property because a table says it has low volatility, the reasoning rests on the reporting frequency of the index that produced the table. If it holds property because rents in a particular local market respond to different forces than corporate earnings do, the reasoning survives however the asset happens to be recorded.
And one sizing point that follows from lesson five rather than from this one. A directly owned building is a large, indivisible, leveraged position in a single object. Whatever a target allocation to this class says, direct ownership tends to overshoot it, because the minimum position size is one entire building and there is no way to hold 60% of one.
Reading a low measured volatility as low risk. The recorded series above showed 5.77% against a true 14.27%, and the worst recorded year was 24.2% of the worst true year. The smoothing is a property of the measurement, and any statistic built on that series inherits it.
Using a smoothed series in an allocation model alongside marked ones. Feeding an appraisal-based property index and a daily equity index into the same optimisation compares two quantities that were not measured the same way, and the method will reliably favour the one that was measured less often.
Concluding that a listed property trust is riskier than a directly owned building because its price moves more. The price moves more because it is quoted more. The buildings underneath are subject to the same conditions, and the visible version is the one telling you the truth sooner.
Relying on the diversification exactly when it matters least. Property is bought with credit, so tightening lending conditions hit it at the same time as everything else. The independence is real in ordinary periods and weakest in the periods diversification is held for.
Allowing one building to set the allocation. The minimum position is one whole indivisible object, frequently leveraged, so a direct purchase routinely takes this class far past whatever the plan said - and the position cannot be trimmed back afterwards.
Write the sentence that says what determines this asset's income and why it is different from what determines your other holdings' income. Local employment, local supply, transport, tenant demand. That argument survives the smoothing problem because it was never a claim about a measured statistic in the first place.
When you meet a volatility or a correlation figure for unlisted property, ask how often the underlying valuations were struck. The demonstration above produced 40.4% of the true volatility from a plausible smoothing, so a figure from an annually appraised index should be treated as a lower bound rather than an estimate.
A listed property trust holding similar buildings is marked continuously, so its volatility is a far less flattered picture of what property does. It is not a perfect proxy - the trust carries its own debt and its own management - but if the two figures are wildly different, the difference is measurement rather than asset.
Decide what share of your total net worth this class should be, write it down, and only then see what that permits. A directly owned building is indivisible and usually leveraged, so it tends to overshoot any target it is measured against - and discovering that after completion leaves no way back that does not cost six per cent.
When deciding how much property a portfolio should hold and in which form, when a published risk or correlation statistic is being used to justify the allocation, and whenever a directly owned building's recorded value has been unchanged for a long period.
Do not use this lesson to appraise a specific building - that is lesson four - and do not treat the smoothing argument as a reason to avoid the class. It is a reason to distrust the statistics, not the asset.
During periods when listed markets are falling sharply, directly owned property frequently shows no change at all for several quarters, and owners take this as evidence that the asset held up. Often no comparable sale has completed nearby, no appraisal has been commissioned, and there is no mechanism by which the recorded figure could have moved. When transactions resume, the recorded value typically moves a long way at once, and the move is experienced as a sudden event rather than as the accumulated arrival of information that had been true for some time. The same sequence produces two entirely different stories depending only on whether anyone was measuring.
A recorded value that has not moved during a period of stress is evidence about the measurement, not about the asset. Ask when the last comparable transaction completed.
Create a free account to read the rest and apply it to your own investing.