Technical Follow Up to Hidden Chinese Debt
There have been some questions posed about some semi-technical issues regarding my last blog post on how large are Chinese debt numbers. Let me note a couple of things before with hit the good stuff. First, regardless of how much we advance the knowledge base, there is vast amounts of unknowns here. We are literally talking about a nearly $40 trillion USD pile that the PBOC dropped into conversation. There is a lot of information that needs to come out about what this means. Second, I am willing to change my mind but at the same time, I telling you what I think based upon what we have been told this means.
1.Is there really a difference between the on and off balance sheet assets? I would argue based upon the evidence we have now that yes, the on and off balance sheet assets refer to two separate assets or pools of assets. I say this for a few reasons. First, the PBOC calls them different pools of assets. The PBOC is most definitely drawing a distinction between the two groups of assets in labeling some as on balance sheet and others off balance sheet.
2.Second, because they are labelled as on and off balance sheet, there is a legal distinction between an on and off balance sheet asset. Using simple examples, if a bank loans money to a company that loan is on balance sheet. However, if that bank arranges an asset management product where lots of investors buy a 90 day fixed income product which channels money to another company, financial or non-financial, the originating bank does not bear the legal requirement to bear that loss. This seems to fit both the requirement of legal difference for an asset to be considered on or off balance sheet but also matches the scant data we have given that roughly 65% of off balance sheet assets are asset management. Now it is unclear whether banks directly hold those assets using accounting rules trickery to ensure they are considered off balance sheet or if banks were acting as the originator and distribution entity and would consequently face significant pressure should defaults occur. Given bank asset holdings, they are likely holding some of this but it would necessitate enormous amounts of onward sales rather than acting as the primary investor. This would also seem to match a point of confusion in the Chinese version of the FSB report where it refers to the “off balance sheet business” rather than assets. Business here might imply that the banks were selling products presumed to have bank backing by investors even if there is not a legal obligation. In all probability, off balance sheet assets here are some combination of both bank owned assets held off balance sheet and bank products sold to investors that banks would be expected to stand behind. This also matches a CBRC document sent to me by Andrew Polk. From Google Translate, the CBRC says this about off balance sheet obligations and products:
Article 2 (Definition of Off-balance-sheet Business) The off-balance sheet business referred to in these Guidelines refers to the business done by a commercial bank that does not include real assets and liabilities in accordance with the current accounting standards, but which can cause the current profit and loss changes
Article 3 (Classification of off-balance-sheet business) According to the off-balance sheet business characteristics and legal relations, off-balance sheet business is divided into guarantee commitments, agency investment and financing services, intermediary services, and other categories.
According to this, banks can engage in off balance sheet activity that matches the two basic types of off balance sheet activity we have defined as the Enron SPV or the investment intermediation. In short, the PBOC is telling us these are two separate asset pools and the CBRC has defined legal distinction between the on and off balance sheet asset.
3.Is it possible that the on and off balance sheet assets are double counting the same assets on both sides? Quite possible to a small extent but that does not change the fundamental conclusion and actually, most likely makes the situation even worse. Let’s walk through an example of how might an asset be counted on both the on and off balance sheet side and how that might actually make it worse before turning to whether or not there is evidence of this happening. Assume for a minute Asset A is held in an off balance sheet entity. For that asset to go from off balance sheet to on balance sheet, that means the on balance sheet entity is incurring a liability. As a real world example, assume a bank sells a wealth management product to investors that gets counted as an off balance sheet liability. The WMP is actually just channeling money into the banks on balance sheet asset base. In this case, moving the money from off balance sheet to on balance sheet creates a liability from the on balance sheet entity to an off balance sheet entity. In other words, this would raise the on balance sheet liabilities if the off to on balance sheet transfer was actually recorded. This would all reverse banks were moving on balance sheet holdings into off balance sheet holdings.
4.On a slight tangent, it would then help to know whether banks are looking to move assets from off balance sheet to on or from on balance sheet to off. This is semi-speculative, anecdote and nothing more, but probably both but in a way designed to make banks look better than they really are. Simple example, bad debts are siphoned off into off balance sheet holdings while capital disguised as deposits is moved on balance sheet. Net result is to make the bank look better.
5.Before we turn to the empirics of whether the assets might be double counted, let us look at why it is almost worse if they are not. Let us assume there is 250 trillion in underlying assets banking system assets. For any number of reasons, now let us assume that the asset is held off balance sheet and circulated on balance sheet (or vice versa). This implies that there is effectively much higher leverage in the banking system than recognized. Take a simple example of how this might work. Assume a bank has 100 RMB in deposits and makes a loan for 90 RMB. They want to make more loans so they turn the loan into a structured WMP and sell it through their asset management division to private investors. That loan is now off their balance sheet and the 90 RMB in cash comes back and they again have 100RMB to lend. If, and this is the key part, if the bank just holds cash on the balance sheet instead of relending the money, there is no net change in risk. If the off balance sheet product collapses the bank can cover the losses. However, if the bank then relends the cash which they have done according to financial data, this means, in our simple example, that there is now another 90 RMB loan made by the bank for total loan assets of 180 RMB (90×2) and 10 RMB in cash lowering the capital reserve ratio if on and off balance sheet assets are counted. The beauty of this explanation is that it matches what little we know about these off balance sheet assets. The scary part is that means that the Chinese banking system leverage is enormous. To provide some perspective, the official capital adequacy ratio for banks is bouncing between 11-11.3%. Now it should be noted for various reasons, the CAR in China excludes a lot of loans made by banks, which Chinese banks know and use these loopholes to boost their CAR. Some research has been done by different people that if off balance sheet items for instance are counted, many small banks especially see their CAR fall dramatically. If I take a simple metric of commercial bank net capital of 15.5 trillion RMB and divide it by on and off balance sheet assets of 485 trillion, this gives the Chinese banking system a net capital to total asset ratio of 3.2%. Should be noted this is not a strict apples to apples comparison. However, it does clearly illustrate what happens if we claim that there is some double counting. One final point about the double counting issue. Let’s assume that all assets overlap or are double counted. Because an asset cannot simultaneously be held off balance and on balance sheet it must be either or, the only way this does not dramatically increase leverage is if the Chinese bank is holding cash offsetting an off balance sheet asset. This would imply that Chinese banks are holding mostly cash or cash like instruments. Well we know that Chinese banks are not holding mostly cash so this leads to the conclusion that Chinese banks have used off balance sheet transactions to further lever up and make their on balance sheet assets appear safer than they really are.
6.This track to find double counting gives us a method to follow the bread crumbs of how we might find evidence of double counting. The primary asset class we are going to focus on are flows to/from banks to other financial institutions or other asset holdings. There is a simple reason for this. Again, take the extreme example that on and off balance sheet assets are the exact same assets. In this case, the on balance sheet financial data should be a record of those assets churn between on and off balance sheet. That would mean that the entirety of official bank data is fraudulent. By that I mean, to take a simple example, the category of lending to household is completely fraudulent because all bank assets are channeled through off balance sheet vehicles prior to consumers. That means all numbers should be recorded differently as being channeled through off balance sheet vehicles and not going to consumers. So the key question then is what is the flow between financial institutions and other financial institutions and or categories that might represent this type of vehicle? For instance, we are going to exclude household consumer bank assets or liabilities assuming that banks are recording loans to consumers as a consumer loan. The primary data source is a PBOC monthly dataset of depository corporations balance sheet. We add up Claims on Other Depository Corporations, Claims on Other Financial Institutions, Claims on Other Resident Sectors, and Other Assets. We then do the same for the corresponding liability line item. According to this, Chinese depository corporations have 106 trillion in assets under these line items but 54 trillion in liabilities for a net holding of 52 trillion. Let’s start with the most generous of parameters by assuming that all 106 trillion in assets here are held in off balance assets so it is effectively double counted. That significantly reduces the 253 trillion we started with to 147 trillion in uncounted off balance sheet assets but that still leaves us with an enormous amount of uncounted assets. Next let’s use the slightly more conservative net asset number of 52 trillion assuming that is entirely double counted assets. This still leaves us with 201 trillion in previously unknown assets. Other datasets which cover financial institutions and depository financial institutions on sources and uses of funds provide smaller corrections, so I will not use those here. If we use the primary line items that would correspond with off balance sheet activities and be very generous in our interpretation, we still are left with a very large amount of uncounted assets.
7.There are a couple of enormous problems with the double counting theory. First, is what I will call the flow mismatch. An asset is categorized on balance based upon where it is deployed. Assume a bank makes a loan to a coal company, that is categorized as a bank asset as a loan to a non-financial corporate. Even if we generously assume all assets from multiple potential line items are deployed as off balance sheet assets, this still leaves us enormously short of double counting even a majority of off balance sheet assets. To claim that all or most all balance sheet assets are simply double counted, you are subsequently required to believe that all on balance sheet financial data is false. Consumer loans should be recorded as loans to consumers. If a bank makes a loan to an off balance sheet SPV that makes loans to consumers, that should be recorded as a loan to a non-bank financial institution or as a portfolio investment depending on how the deal is structured. Remember, roughly 65% of the off balance sheet assets are held in asset management structures which is not how Chinese banks record holding their assets. These do not match. The balance sheet flows and categorizations simply do not come close to matching. Second, is what I will call the size problem. The only way the double counting theory makes a significant difference is if we assume effectively that all on balance sheet banking assets somehow move through off balance sheet banking channels before reaching their final destinations. This is also the only scenario I can think of that doesn’t drastically raise the risk level. In this instance, Bank A makes a loan to Bank A SPV who makes the loan to the end customer. If the off balance sheet SPV makes loans to consumers, the on balance sheet entity records the loan as being made to consumers rather then to a non-bank financial institutions or portfolio (WMP) investment. Is this possible? Given it is China we are talking about who just disclosed nearly $40 trillion in previously undisclosed assets, anything is possible. However, this has never been discussed even anecdotally, requires us to believe all on balance sheet financial data is wrong, and that the entire Chinese banking system is engaged in a systematic asset obfuscation and diversion scheme. Possible? Sure. Highest probability explanation? Not even close. Therefore, if we take the Chinese banking data we have, believe consumer loans are made to consumers and so on, even if all asset classes we can remotely presume to be in off balance sheet vehicles are in off balance sheet vehicles, we simply do not come close to reconciling the outstanding unexplained assets. I am quite willing to believe there is some immaterial level of overlap here. For instance, assume 10% of the off balance sheets are already counted on balance sheet that would reduce the unknown by roughly 25 trillion RMB (which let’s just stop right there and say that is still an enormous number) to about 225 trillion. 225 trillion RMB or $34 trillion USD is still an enormous amount of unexplained assets. Based upon all the data we have, it seems highly unlikely that a large majority of the on and off balance sheet assets are simply double counted.
8.How was China’s last figure of financial system assets totaling 833% of GDP estimated? The FSB gave the total financial system assets for China across Central Bank, Banks, Insurance, Pension, Public Financial Institutions, and Other Financial Intermediaries at the end of 2015. To estimate the financial system assets as a percentage of GDP at the end of 2016 with the new PBOC data required the following steps. 1) change the FSB 2015 bank asset to 2016 PBOC on and off balance sheet asset total 2) total PBOC assets at end of 2016 3) Estimate 2016 growth rates for asset growth rates like insurance using conservative growth rates of 10-12%. Insurance for instance grew at 22%. (Worth noting inserting PBOC data from on and off balance sheet asset into FSB table comprises 80% of financial system assets). 4) Sum estimated total financial system assets for 2016 from FSB with new PBOC data and divide by IMF total nominal GDP.
Let me emphasize, and a couple of people have the DMs to prove it, when I first saw these numbers I simply did not believe it because the numbers were so outlandish I thought I had to be missing something. I am still open to changing my mind on this issue. However, the PBOC and the CBRC both appear to be drawing a clear statistical and regulatory dividing line between on and off balance sheet assets. Furthermore, the asset flows between on and off balance sheet entities simply do not match either in asset categorization or amount. To believe that the on and off balance sheet asset values double count the same assets means disregarding CBRC regulation, PBOC classification, and all on balance sheet banking system data. It is worth reminding that the PBOC FSR in previous years mentioned the ongoing build up of off balance sheet assets. In 2016 it amounted to 82.36 trillion and in 2015 it was 70.44 trillion. 2017 changed because of the inclusion of the MPA.
Finally, I think there are so many questions that need to be answered with regards to this disclosure. I think it clearly says that is roughly $40 trillion USD in previously undisclosed assets which is nothing short of a complete game changer on everything.