A missing trillion? Is the total value of UK property understated?

The Statistics

We’re likely all familiar with two headline statistics which come out each year.

The first is the average price paid of a residential property in the UK, the second is the total value of all properties in the UK. The former is normally released by the ONS and the latter is usually announced by Savills, with both then being used for temporary headlines.

Source: The Guardian

Source: The Guardian

But have you ever noticed how they don’t add up?

Depending on what measure you use, there are 26, 25, or 22 million residential dwellings in the UK, and all kinds of variations in between.

Now if you take that average house price of £291,699 and multiply it by 26, 25 or 22 million, you will not get to £8.7trn. You’ll get at most £7.6trn. There’s a trillion missing!

Why The Difference?

I’m sure you’ve already worked out why. This is because the total value of properties in the UK is calculated from the average estimated value of each individual property, not the average sales price, and the average value of properties in the UK must therefore be higher than the average sales price.

Well, how much higher? As you can see from the extract below, our systems currently value the average property across the UK at £350,549 which is about 20% more than the average prices being paid (£291,699).

Now two questions emerge. Firstly, why are they different? Surely the average price being paid for an average property across the UK should be the same as the average value of the average property in the UK? And secondly, how are these values determined and how reliable are they?

First question first, and the answer is actually in the question itself; ‘an average property’, the thing is, the average property being sold today, is not the same as the average property being owned today. There is a difference between the 22m-25m homes that exist in the UK which aren’t being sold each year, and the circa 1m homes which are being sold each year.

Take for example, and for no other reason than I happened to click on it first, the East Riding of Yorkshire. Semi detached properties are the most common type (37%) but they are not the most common type to be sold, they’re actually in second place (34.5%) compared to Detached properties (36%). So semi-detached properties are underselling relative to how common they are. On the other side, Terraced properties make up 19.3% of the housing stock, but 23.4% of the sales, so they are selling more relative to how common they are.

The average price paid for a Semi-detached property is £194,000 and for a Terraced property it’s £153,000 so just from this simple calculation, you can see how the increased sales rates of Terraced properties would bring the average price paid down, compared to the semi-detached properties which are not being sold. If they were being sold in proportion to their portion of the housing stock, we would be expected to achieve a higher price paid. In a later article we’ll look at this more closely as well as the generational and economic factors at play, but in short, the type of properties being bought and sold are not exactly representative of the housing stock, and are generally cheaper than the overall housing stock, which explains the difference between the average value of properties and the average prices being paid.

How Reliable Are Valuations Compared to Prices?

Next is the question of how are these valuation estimates determined and how reliable are they? A transaction is quite reliable, it clearly happened, a buyer and a seller exchanged ownership and some money. But the value assumptions for properties that are not sold are simply that, assumptions.

There are numerous ways which a property can be valued automatically; hedonic models, sales indices and comparable analysis to name the most prevalent and well considered methods. While I can’t speak for Savills methodology, their figures do align with our own which mixes elements of the three methods just mentioned. In consideration of the accuracy of the models being used, there are two factors, both the accuracy of the model on any individual property and the accuracy overall of all the valuations brought together. In the first case, accuracy can vary, while most ‘normal’ properties can be valued quite easily and accurately using an AVM, more unusual properties, which are inherently rarer, are harder to value. Generally speaking, these rarer properties are a faction of a percent of the overall stock and so don’t make much difference to the overall figures.

Then in regard to the overall accuracy, in this case we benefit from an aspect of the law of large numbers and the innate requirement of any valuation model to remove biases. Effectively any model should not have an overall bias for over or under estimation, it can get the valuation wrong, but it shouldn’t be consistently wrong in one direction. Once that is dealt with by tuning the model, the overall bias should be at or near zero, and then the net result of any slight overvaluation on one property, is counteracted by the equally likely undervaluation of another property.

Then there is the question of reliability. The models are set to attempt to predict the sales price of an arm’s length transaction in the market. With circa one million transactions per year and around twenty-five million properties existing, the annual transaction rate is only about 4%, which is many multiples lower than a publicly traded stock (by my back of a napkin calculation, 81% of Apple stock is traded each year). This isn’t news for anyone, real estate is relatively illiquid, however there are plenty of small cap stocks which trade with similar transaction rates without problems. Provided that there is no sudden reason why arm’s length transactions couldn’t take place then the amount of liquidity in the market is fine, and even if it were to briefly stop, given the length of history of information saying that the price or value of a property is X, it would be unlikely that prices would suddenly collapse in a few days. There is a caveat to that, which was an issue discovered during the last financial crises, that if everyone is trying to sell, and no-one is wanting to buy, then the prices of assets will collapse to zero regardless of the theoretical value computed by various models. In theory it could happen to real estate, however, unlike financial instruments, real estate is a necessity and so it is very unlikely.

Conclusion

In conclusion, the figures on the overall valuation of property in the UK do appear to be correct at around £8.7trn, and the difference between the average value of a property in the UK and the average sales price of a property in the UK will be covered in a follow-up article.