Is Chinese Mortgage Data Waaaay To Low? (No, seriously)

So recently a lot of ink has been spilled on the rapid growth in Chinese mortgages.  On the face of it the increase is certainly worrying.  New mortgage lending in 2016 is up 111% and the total stock of mortgages is up 31%.  Even if we take a broader measure of household lending that likely captures a not insignificant amount of real estate related debt, medium and long term loans to households is up 31%.  The numbers on their face appear large with medium and long term loans to household registering 22 trillion RMB and personal mortgages clocking in at 16.5 trillion RMB.

These sound like big number and in some ways they are, but in reality these numbers are if anything suspiciously too low.  Most get caught up on the size of the numbers but never place these total numbers in any type of context.  In fact, if you place these numbers in context, these numbers are absurdly low.  Let me explain.

For conservatism, data, and simplicity sake, I am going to limit the analysis to urban housing units.  In other words, let us assume that all mortgage and medium to long term household debt is owed only by urban households.  This does not change the outcome in anyway and if anything make it much more conservative than it would be otherwise.

The primary thing we want to do is adjust for the number of households in urban China.  Without going into all the underlying calculations, which come from all official data, there are approximately 272 million urban households in China and according to official data, only a very small number of households do not own their housing.  Again, this is all relying and strictly using official data.

If we then estimate urban residential real estate wealth using the 100 City Index price per square meter as our high value and the Third Tier City Price per square meter as our low value, we have both a high and low value for our estimate of urban residential real estate wealth.  This gives us an estimated upper bound of 330 trillion RMB and a lower range of 189 trillion RMB.

Here is where it gets interesting.  If we translate this into a broad loan to value number, this means that urban China has an estimate loan to value ratio on its real estate holdings of 5-9%.  In other words, almost all of urban Chinese real estate is owned almost entirely free and clear according to official statistics.

If we apply this analysis backwards, the numbers are even more nonsensical.  In 2011, the urban loan to value ratio ranged from 3.3-4.5%.  If we use absolute numbers, the appear even more absurd.  When the average housing unit in 2011 cost 665,000 RMB using the third tier city price and 910,067 using the the 100 City National Index, mortgage debt totaled only 29,675 RMB per urban housing unit.

If we focus just on the new mortgages and new urban units, the numbers look decidedly problematic.  For instance, if we use the 100 City Index housing price, this would give us an implied equity share for new housing units from new mortgages of 71%.  In other words, if we assume that only newly constructed units are purchased with new mortgage debt, owners would be providing a down payment equal to about 71%.

Now while I use the slightly more restrictive mortgage debt, even if we include the broader label of medium and long term this would barely dent the number.  If we use the medium and long term household debt number instead which is only about 4-5 trillion RMB more, again using only urban households, this would still barely move the per unit or value debt number.   To bring Chinese urban housing wealth up to a 20% LTV, would require about a 41 trillion RMB increase in mortgage debt.  Put another way, outstanding mortgage debt would need to go from about 16.5 trillion RMB to 58 trillion RMB. Including the obvious candidates that some have nominated simply does not come close to making these numbers plausible.

We are left with a conundrum: either believe the data at these levels or find a better candidate when no good obvious source of debt under counting exists.  I’ll be honest in saying I’m not sure whether to accept them as vaguely reasonable representation or believe that they are not even close.

If we consider the possibility that these debt numbers are relatively accurate, while there are positives, there are also very real risks.  First, it raises the scope that Beijing could further increase urbanization and home ownership rates by loosening credit.  However, there is evidence that rural households migrating to urban areas are already debt budget constrained and that Beijing is uncomfortable with the level of debt even at these levels.  Additionally, this raises the possibility that real estate prices have a long way further to appreciate which seems implausible given already elevated price to income levels.

Second, this would imply that households have put very high level of savings into their homes and may have less liquidity available than understood.  By some recent estimates, Chinese households had 70% of their wealth in real estate.  Liquidity constraints may exacerbate any real estate or broader economic down turn placing additional pressure on prices.

Third, this would seem to place enormous pressure on public officials to maintain housing prices at elevated levels.  If Chinese households have placed the vast majority of their wealth into their home, though lack of leverage will not magnify the financial returns, it will place enormous pressure on the government to prevent price declines.

There is one possible scenario, though we do not have the data to say for sure this happening that would explain the discrepancies we see.  Given the mismatch of the mortgage data and required down payment this raises the possibility of the leverage upon leverage scenario.  For instance, a home is owned with no mortgage debt.  The owner then pledges the real estate as collateral to borrow money for the equity share and borrows money in the form of a mortgage to purchase additional real estate.  In this instance, only one mortgage appears outstanding where, if we assume the second property is financed with a 50/50 debt/equity split at the same value of the first property, then we have a mortgage per unit value of 25%.  However, in reality the risk level is much higher as both properties have debt against them and depend on stretched cash flow valuations or capital appreciation.

There are many possibilities but the only thing we can say for sure at the moment, once we break down mortgage data into per housing unit basis, the numbers seem implausibly low.


10 thoughts on “Is Chinese Mortgage Data Waaaay To Low? (No, seriously)

  1. I think that maybe there are a couple of points worth bearing in mind. First, it may be misleading to focus on the tier 3 average price and the top 100 cities average price for property as I believe the majority of Chinese people and combined urban income derives from tier 4 and 5 cities. Assuming i’m right, you would significantly reduce the value of property using tier 4 and 5 numbers and thus increase the loan to value ratio.

    Secondly, based on conversations with Mainland colleagues, a large percentage of homeowners in china were simply given their homes by their state-owned enterprise employers over the past 20 years. These (often older) urban residents have very low incomes, but live in homes which have hugely appreciated in value. As an aside, this is one of the main reasons why China is struggling to introduce a property tax, despite the obvious advantages of doing so. These older, ex-SOE employee owners are the bedrock of party support, but simply could not pay a property tax based on the value of their property.

    so i think you may be using too high a value for your average property and I think you may find that a large percentage of owners have no mortgage (because they were given their home) and this will then push up the loan to value ratio for the owners who actually had to buy their properties.

    As you say, this may not fully explain the numbers, but I suspect it would go further in helping to understand the raw stats.

    • 1. On the the gifted apartment, you are totally right here is where I get skeptical. The levels are so low as to have required virtually the entirety of Chinese urban housing stock to have been gifted with virtually no new buying. That is such an extreme number as be questionable. I’m definitely open that the wealth transfer was just that large and the new buying/turnover was so small but right now we just don’t have the ability to demonstrate that with data.

      2. I’ll have to look at what percentage of the population lives in tier 4/5 cities vs. other cities. Most data only goes down to Tier 3. I suspect it isn’t as big as you think but very willing to consider it as a possibility.

      • If it helps:

        According to Nielsen’s Winning in China, Insights and Strategies for 2011 (

        Tier 1 Cities: 16 million households, 1 trillion income value
        Tier 2 Cities: 38 million households, 2 trillion income value
        Tier 3 Cities: 75 million households, 3 trilllion income value
        Tier 4 Cities: 86 million households, 3 trilllion income value
        Tier 5 Cities: 169 million households, 4 trilllion income value
        So, assuming this is correct (and I’ve seen other breakdowns along the same lines), then clearly the vast majority of chinese households (albeit the poorer ones) live in tier 4 and tier 5 – as such the value of property in those is probably more relevant. but as you say, this may not explain the whole discrepancy.

  2. You should pay attention to the structure of the Chinese housing market. According to the 2010 Population Census data, only about 36% of the urban households purchased their homes from various sources (new homes, existing homes and low-cost housing from the government). In the rural areas, most homes are self-build (94% according to the census data).

    Reference (webpages in Chinese):

  3. Dear Professor,

    thank you for your insight. Is it possible with your available data to show how LTV of 1) total market and 2)of new mortgages and new urban units, develop over time? You would then be able to compare the explanatory power of the price growth and of the debt availability growth.

    • Let me see if I can answer that if I understand it correctly: the yearly change in mortgage lending relative to new housing units has pushed yearly LTV higher. By that that I mean this, new urban housing units has fluctuated within a band of 8-11m a year for a number of years. However, new mortgage lending has gone up rapidly. So if we assume that new mortgage lending is only flowing to new housing units, a little unrealistic but not entirely for reasons I’ll explain in an upcoming post, that implies that the LTV for new housing units is moving away for this other number and relatively quickly. I’ll post a table about that when I get a chance.

Comments are closed.