Needless to say that there is a huge bubble in Chinese real estate, just as Japan had twenty years ago. While it is almost ludicrous to say that there is no bubble, because of the poor quality of data (or sometimes, no data), it is difficult to tell exactly how big the bubble is.
One of the approaches we can take is casual observation, as I have already done for the extreme case of Hainan. And indeed, I generally feel that the problems of having large numbers of unoccupied buildings exist in almost every city in China. One obvious weakness of casual observation, however, is that it is very hard, if not impossible, to generalise your findings in such a big country.
If we look at the data instead, I am just as frustrated as everyone else because of the lack of reliable data. There are, of course, some pieces of data which consistently offered the impression that real estate investment increased significantly after the 2008 financial crisis as the Chinese government pumped in massive stimulus. For instance, we did see new constructions boom across the board towards the end of 2009 and 2010 after contracting in the first half of 2009:
But this is almost where consistency ends with official data. From various sources of official data with various calculations we can establish, albeit in a convoluted way, that potential housing supply at the national level (i.e. existing housing stock plus those already under construction and those land that has been sold to developers but constructions have not yet been started) is far too high for the entire population. The existence of oversupply is a question of how much.
In determining the future potential housing stock, the starting point is the extent housing stock. In this, I have largely got to the same starting point as Jinsong Du of Credit Suisse has. First, I took the reported Per capita urban residential GFA (which are different numbers from using the total housing stock divided by reported urban population), multiplied that by the non-agricultural population (hukou basis) to arrive at the existing housing stock for 2010 at 14,514 million square meters. On that basis, the annual net increase of urban housing on average was 816 million square meter per year from 2005 to 2010. For simplicity sake, the 2011 existing housing stock is estimated by adding the 2010 housing stock and the annual completion of residential floor area. This is not a completely accurate way of doing so, as we would normally have to account for, among other things, the floor areas being demolished, etc, which should point to a lower net increase in GFA. However, the figures show that the annual completion (per National Bureau of Statistics) is invariably lower than the net increase, thus for the purpose of this exercise, the reported 717 million square meter completion is used.
With respect to future supply, one can base the following analysis on the 2011 third quarter report from the Ministry of Land and Resources, which stated that the land area for residential uses which was under construction amounted to 355,000 hectares, and land area for residential uses of which construction has not started amounted to 165,000 hectares. Note that there are later figures, which said that land area under construction by the end of 2011 amounted to 479.9 thousand hectares, but as that did not separate the portion for residential use, I’m using a smaller and less updated figures of 355,000 thousand hectares. That means the potential residential GFA supply currently in the pipeline will come from the total of 520,000 hectares.
Note that these figures are in terms of land area, not actual floor area. Thus to get the potential housing stock, we must multiply the land area by an appropriate plot ratio. It is difficult to say for certain which plot ratio to use. Some suggests that 1.5x is the right ratio, while Jinsong Du used 2.2x. Below I have estimated the potential housing stock on various plot ratio assumptions, from the low of 1.5x to an unrealistically high ratio of 3.0x:
The resulting total housing if all the land is developed under different plot ratios assumptions will be as follows:
Now, when it comes to the demand side of the equation, instead of trying to classify various types of annual demand (e.g. first-time buyers, upgrade buyers, investment, etc), which is a flow concept (similarly, I focus on estimating the total housing stock in the future rather than annual new supply), I will focus on the population growth and the rate of urbanisation as the main factors, which should give me the indication of total floor area required in the future.
The reason I focus on the stock concept of total housing stock and total floor area required as implied by total population in the future and rate of urbanisation is that real estate stays there after being bought, used, and subsequently abandoned by the owner (to be sold to others). Or to put it more clearly, upgrade buyers will most probably sell their existing homes, which should be subsequently on the market, and people who buy real estate as an investment will (eventually) have to sell it to someone else ultimately to realise on profits (if any), or let it out at the very least. That makes the analysis that focuses on, for example, annual investment demand vs. annual completion totally useless, because even though investment demand can absorb all completion in a given year, this real estate will eventually appear on the secondary market, and that will (if using the flow concept) add to the “supply” in the future years.
For the forecast of total population, I use the forecasts made by the UN population division. In terms of the future rate of urbanisation, the various scenarios are below:
Additional assumptions regarding per capita GFA have to be made. As mentioned before, the per capita GFA officially reported by the government does not match the figures derived by dividing the total housing stock by total urban population, with the total housing stock being reported by the government until 2006. As a result, the reported per capita GFA appears to have been overstated.
In estimating the total housing stock required, I use the derived per capita GFA as a starting point (which is invariably smaller than the reported figure, as state above), and assumed that it will simply go up to 30 meters per person by the end of 2020. With all these estimates (population, urbanisation rate, per capita GFA), rough estimates of total GFA needed can be determined. And the result indicates that by the end of 2020, under various assumptions, the total GFA required around 24 billion square meters to around 27 billion square meters:
Using the most bullish assumptions for both plot ratio (at 1.5x) and urbanisation rate (65% by the end of 2020), one can conclude that if developers completed all the constructions on all the lands which are currently under construction and those which construction has not yet started, the resulting floor area could easily accommodate all the residential needs for the urban population at least through 2017. Using a moderate assumptions of plot ratio at 2.2x and urbanisation rate at 58.5% there is sufficient floor space to the end of 2020.
Note that the population forecast from UN population division suggests that China’s population will peak in 2025 and decline thereafter. If that is indeed the case, and if the current potential residential supply in the pipeline can indeed meet the requirement through to 2020s, China will only need to complete all the constructions on these 520 thousand hectares of lands and stop doing any constructions whatsoever, and China will be able to meet the residential requirement pretty much forever:
Of course, one has to recognise that this cannot be a precise estimate. However, this analysis does establish (for me at least) that it is not an exaggeration to assert that the housing supply in the pipeline is sufficient. Certainly, it is not likely that construction will stop altogether, nor are we sure that all those currently in the pipeline will be completed. For example, we can imagine that some developers will run out of money and be unable to complete the work. Also, it is difficult to quantify the exact amount of residential floor area that will be demolished over the next decade, and that could make the current estimate of future housing stock an overstatement.
Another source of vagueness comes from the starting point of the whole exercise. The existing housing stock is one of the big uncertainties here. Nobody is exactly sure about the occupancy of the existing housing stock. We have already seen that large number of flats are empty as people bought for investment, not for their own use. But this stock will have to be sold to real users ultimately, or at least be let out, in order for investors to realise any return. it could take extra years just to fill up these vacant homes, which may further intensify the problem of oversupply.
A final source of uncertainty comes from the rate of urbanisation. Although it has been progressing at a very rapid rate over the past decades, urbanisation in China still lags major developed countries. One could argue that urbanisation will increase the requirement for housing. However, as Jinsong Du points out, at least some of the urbanisation was happening simply by re-designate some areas from rural to urban, and people who have already been living there already owned houses. As a result, that does not increase demand, but it adds to housing stock. Secondly, I doubt if Chinese urbanisation will catch up with other developed economies. Research has pointed out that agricultural land as a proportion of the total land surface of China is well below the international average. Urbanisation not only reduces agricultural land through development but those lands not being developed are affected by urbanisation through pollution and other effects. The food security of China may pose a limit on rapid urbanisation.
Real estate investment now probably accounts for between 10% and 13% of China’s GDP. With over-investment in fixed assets (not only in real estate) in various parts of the economy, it is inevitable that investment will slow down. As a component of that, the above mentioned residential supply will all be completed by 2015, and that supply can meet demand effectively forever. It is preposterous to suggest that construction will stop on the land surface of China, but construction will inevitably slow down in the years to come.