My Blog

February 13, 2008

Yesterday at an investor meeting someone made the point that the lack of transparency in Chinese accounting may actually act to reduce the riskiness of the system.  If this is true, it is pretty good news for investors banking on government credit. 

What’s the total amount of central government debt in China and how is it structured?  We don’t really know.  What is the amount of municipal and provincial obligations guaranteed by the central government?  We certainly don’t know.  How much did banks lend against real estate?  We have some figures but it is widely believed that an awful lot of real estate loans are not correctly classified as real estate loans.  What is the breakdown within the banking system of new loans, and what about old non-performing loans?  We have some information but not nearly enough to give us a real sense of how much bad debt there currently is and, perhaps more importantly, how vulnerable new loans are to a slowdown.  There is a lot of mystery embedded in the national balance sheet.

Asking about the creditworthiness of the central government is not a trivial question.  China is in a reasonably good fiscal position and government expenditures only barely exceed revenues.  In addition most estimates put total current debt of the government at around 30% of GDP, which everyone agrees is fairly low and should cause little concern to anyone looking at the government’s ability to service debt. 

More importantly from a financial stability point of view, most government obligations are very safe and optimally structured – they consist of medium- and long-term, fixed-rate, local currency bonds.  In my book on financial vulnerability in developing countries I make a point about stressing how important this kind of debt is in protecting countries from the threat of financial crisis because of the way they dissipate the impact of adverse shocks.  In a crisis anything that causes interest rates to rise substantially (say a burst of inflation) would automatically reduce the existing debt burden, thus acting to stabilize the financial system.

But still, that doesn’t mean that there is nothing to worry about.  Last March the Inter-American Development Bank published Living with Debt - How to Limit the Risks of Sovereign Finance by Eduardo Borensztein, Eduardo Levi, and Ugo Panizza, and one of the most interesting points the book makes is that the surge in debt that precedes a financial crisis rarely occurs because of the accumulation of massive fiscal deficits, as we are likely to assume, but rather because of a sudden conversion of contingent liabilities onto the government’s balance sheet. 

The two most likely sources of these contingent liabilities have typically been unhedged external debt, when a rapid depreciation of the currency suddenly causes the value of external debt to rise relative to the value of domestic assets, or a surge in non-performing loans on bank balance sheets that forces the government to intervene and assume the liabilities of the banking system.  China, which has been vigorously fighting the 1997-Asian-Crisis war, is almost immune to the former risk, but it doesn’t take an alarmist to see that the latter risk is a real possibility, especially given the explosion in bank lending during the best-of-times period of 2003 to the present.

Right now, my best guess is that if various contingent liabilities were correctly accounted, total liabilities of the government would exceed 50% of GDP, and could be much larger – some estimates put it at 80% of GDP, which is not implausible.  If there were a significant downturn the number would probably get worse – it is almost a certainty that non-performing loans in the banking system would rise in an economic downturn.  How much is anyone’s guess, but it is not implausible to assume the possibility of a sharp rise in non-performing loans. 

When these are combined with the non-performing loans still residing on the bank balance sheets (some correctly identified, but perhaps many of them still hidden) and the loans transferred onto the asset management companies created for that purpose, who are technically bankrupt but explicitly guaranteed by the MoF, total debt of the government may jump substantially, and this debt is not nearly as well structured as the existing bond obligations of the government.  In fact, as I explain in my book and elsewhere, these types of debt exacerbate external shocks and so can be as toxic for the government as external debt, or Mexico’s famous Tesobons, in case of a domestic crisis, because like external debt their value tends to expand just when the country can least afford it.

I brought this up yesterday at an investor meeting and received a comment, in the form of a question, that I have heard many, many times before – and from some fairly sophisticated sources.  Since the government has been able to hide the true extent of its liabilities quite well, why should we assume that in a contraction we will suddenly get access to this information and, if we don’t ever know, why should it matter?  The hole on a borrower’s balance sheet is only a problem if we know it is there.  China’s lack of transparency actually reduces the risk of bad information spooking investors.

It is always frustrating to get this kind of comment, especially now when the sub-prime crisis has made it very obvious that sometimes the lack of information is worse than unhappy information.  After all it wasn’t the extent of the potential sub-prime losses, even assuming the worst possible case, that scared the market but the fact that no one knew where these losses were hiding.

Information does matter, especially when we need it most.  Based on many years of trading, I can say one of the few rules of financial markets I know is that when markets are buoyant and liquidity plentiful, transparency and high-quality information is of little use – in fact since we tend to assume the best, no information just means no bad information, so buy, buy, buy.  However when things go badly, investors change tack suddenly and begin to assume the worst.  In this case no information means nothing but bad information.

It’s always dangerous to make predictions but one thing I will predict with great confidence is that when things turn badly in China – when economic or monetary conditions are contracting – the lack of information that emboldens investors today will force them into panic selling.  Lack of transparency is only a blessing in the irrational state of a bull market.  However it becomes a source of new irrationality in a declining market. 

This, by the way, is not a new observation at all.  On my flight to New York I was reading Russell Napier’s Anatomy of the Bear and came across this June 15, 1932 quote from the Wall Street Journal: “During the roaring days of the bull market, lack of full information about a company gave its securities a certain mystery value.  The long depression has done a good deal to eliminate mystery value from consideration of the worth of a security.”

Comments (17) for "China’s mystery value does N...

As an investor I would be completely skeptical about your claim to have heard real investors say that China's lack of transparency would be a good thing in bad markets if I hadn't heard the same thing so many times myself. It is wonderful how willingly we delude ourselves. By TR - 2/13/2008 6:27 AM

I don't think that the numbers are that mysterious. First the total amount of central government bonds issued by the Ministry of Finance is a known number. The amount of provincial and local debt you can guess at by taking a look at the total amount of provincial and local spending, and my back of the envelope calculations says that it is about $100 billion. The amount of NPL's in the asset management companies are also in the $100-$200 billion range. We know the totally amount transfered from the banks to the AMC's. We don't know the recovery rate, as a worst case scenario we can assume that everything in the AMC's is unrecoverable, which isn't too far from the truth. Real estate we can get from the total amount of loans issued multiplied by an expected default rate. Contingient liabilities on corporations can be calculated from a social welfare standpoint. Basically, the PRC government is not going to bailout out your company unless it thinks that you are going to end up with lots of angry people on the street. So you can guessestimate the amount of corporate liabilities by figuring out how much much the PRC government will have to pay out to keep people happy. I can't get any numbers more than $100 billion. Putting all of the numbers in a spreadsheet, my conclusion is that you'd have to put in extremely high and in my opinion implausible numbers in order to get anywhere near a crisis. One other point. People talk about the "lack of transparency" in China as if Hu Jintao has a secret closet somewhere where all these numbers are known. In reality, when we don't know, it's because no one knows. For example, the badness of a real estate crash depends on how many non real-estate loans are actually real-estate loans, and that is a number that the banks don't know, since if they knew about a hidden real estate loan, it wouldn't be a hidden real estate loan. Future non-performing loans have the same problem. All loans are performing until they are non-performing. The other thing that you have to be careful about is double counting. For example if you just add together hidden local government liabilities with hidden corporate liabilities with hidden real estate liabilities, you get a meaningless number since you have loans to local government owned corporations to buy real estate. By Twofish - 2/13/2008 8:44 AM

I think you are completely missing the point, Twofish. You can estimate all you want, but your estimates are likely to be heavily afffected by market conditions, and those estimates will exacerbate conditons at the time because they will underestimate risk when conditions are good and overestimate risk when times are bad. Tha is the point. Your estimates, by the way, are a ittle glib and pretty worthless. For example, you simply assume that total central government debt is the sum of outstanding bonds. Why would you make that assumption? I don't know anyone else who would. There is almost certainly more debt (the government has always had many other forms of debt), but we have no information on how much. I am pretty sure that in case of trouble rational investors will assume there is a lot of guarantees and loans that we don't know about. By the way, you've made what no one has been able to do sound very easy. Adding up all the debt has kept analysts very busy and in very little agreement. For example, you say multiply the amount of loans collaterlaized by RE by an expected default rate. Setting aside that your expected default rate is a meaningless estimate (under what economic conditions? 11.6% GDP growth or collapsing real estate prices? -- if the latter it is likely to be an extremely high number), the amount of RE loans is unknown precisley because there is so much anecdotal evidence that a significant portion of loans collateralized by RE are not registered as such. An example of how uncertain your estimates are is your $100 billion estimate for provincial and municipal debt. I am not sure how you added this up, but a Chinese economist in a well-known piece two years ago estimated that just the unrecorded and uncollectible portion of provincial debt was over $200 billion. That suggests that total debt must be somewhat higher than $100 billion. In a way you sort of demonstrate my point about how subjective these estimates are and how affected they are by our own sentiments at the time. By the way why is it relevant that Hu Jintao doesn't have the numbers in his closet somewhere? Are you arguing that George Bush knows where all the sub-prime losses are and is not telling us, and that is why the crisis spread? I am not sure why a lack of information that isn't in a secret file in the president's office is the only lack of information that we need to worry about. By Michael Pettis - 2/13/2008 9:21 AM

Some notes. 1st, "What is the amount of municipal and provincial obligations guaranteed by the central government? " What do you mean on municipal and provincial obligations? Correct me if I am wrong, I think the municipal and provincial governments are not allowed to issue bonds in China. 2nd, I am agree that information does matter. And I am all for it. But the difference between China and current credit crisis is: There is only one player undertaking those bad debt in China: Chinese gov. It does not matter which agency actually has the problem. People can always look at the aggregate level data to figure out whether the bad debt is sustainable. Thus, creditworthiness is derived from only 1 player in the market. It eliminates at least 1 uncertainty on cherry picking. In U.S., you have no idea who is the "lucky" one in several thousand financial firms. Although the aggregate amount of bad debt is not a big deal compared with the whole U.S. economy. Individual investors can potentially be wiped out if they got "lucky". By fatbrick - 2/13/2008 9:35 AM

the most worrisome part of national debt is the implicit -explicit pension , medical liability government assumed. The pay as you go pension regime and medical care system is reported underfunded by10-20% of GDP, now the coverage is not complete in urban areas, as labour forces peaked out in 2010, this could swell to 50% of GDP very quickly the level of coverage is not rational, now a 60 year old retiree in major cities already get more pension payment than a college graduate salary, By isaac - 2/13/2008 4:16 PM

Pettis: You can estimate all you want, but your estimates are likely to be heavily afffected by market conditions, and those estimates will exacerbate conditons at the time because they will underestimate risk when conditions are good and overestimate risk when times are bad. That's not necessarily the case. There's no reason that I have to make optimistic assumptions. I can make pessimistic assumptions. The purpose of the exercise is to see how pessimistic the assumptions have to be before we have a real problem, and to see what numbers we have to know in order to figure out what is going on. In a lot of organizations, there are institutional pressures to come up with optimistic numbers, but there are also sometimes pressures in other contexts to come up with numbers which are overly pessimistic. One way of removing the institutional bias toward optimism is to change the question. Instead of asking whether or not China is in good or bad shape, ask the question "what assumptions have to exist for China to be in good shape?" and "what assumptions have to exist for China to be in bad shape?" Pettis: For example, you simply assume that total central government debt is the sum of outstanding bonds. Why would you make that assumption? Debt with guarantees that we don't know about isn't guaranteed debt. You start with the debt that the government has a legal commitment to honor. This includes only central government bonds issued by the Ministry of Finance. Then you go into debt that has a quasi-guarantee. There are limits to how much you can hide these, since there are limits to how much you can make a credible guarantee without anyone knowing. It's possible that Hu Jintao has promised a trillion dollars to someone, but good luck on enforcing that agreement. Pettis: There is almost certainly more debt (the government has always had many other forms of debt), but we have no information on how much. We can make a list of all of the liabilities that the Chinese government could have. "No information" is a vast overstatement. There are a limited number of ways that you can convince people to give you money. Also one way of tracking things is with total spending. If someone gets $50 billion, they presumably has to spend it somewhere. Pettis: I am pretty sure that in case of trouble rational investors will assume there is a lot of guarantees and loans that we don't know about. And I'm sure that when push comes to shove, that a lot of these guarantees and loans will turn out to be defaulted. Also, I'm sure that there are lots of deals and guarantees, however it's really hard even in China to borrow $5 billion without someone noticing, and anything less with less than $5 billion aggregate value we can ignore. Pettis: the amount of RE loans is unknown precisley because there is so much anecdotal evidence that a significant portion of loans collateralized by RE are not registered as such. So lets guess a number. 5%? 10%? 100%? Put the numbers in the spreadsheet for each set of assumptions and see what comes out. It may turn out that this number is extremely unimportant. It may turn out that that number is extremely important. If that number is important then we try to lower the bounds. Pettis: A Chinese economist in a well-known piece two years ago estimated that just the unrecorded and uncollectible portion of provincial debt was over $200 billion. That suggests that total debt must be somewhat higher than $100 billion. In a way you sort of demonstrate my point about how subjective these estimates are and how affected they are by our own sentiments at the time. It depends on what he defines as unrecorded and uncollectable, and what he defines as provincial debt. Is he including rural credit cooperatives? Tax arrears by township village enterprises? Unfunded pension liabilities? Depending on what his definitions are, his number might be high or low, and I might agree or disagree. The number I got was to look at the total amount of local government spending and then multiply it by a random guess as to the amount due to loans. It's all a guestimate, but since I've stated where I got them numbers from, you can do statistics to get a better number. If you have me guess a single number, then yes, it's going to be very subjective. If you have me do an itemized breakdown into fifty categories, then the subjectivity decreases. Also single numbers are useless for risk managers, the thing that you really want to know is an itemized breakdown so that you can manage your risk. The other thing that you get if you have an exhaustive list of places where you might have a liability is that if something strange happens, you know how much to pay attention to it. You check your list of possible liabilities, and if you catch any suspicion that you are missing something from that list, you pay really, really close attention to that. Pettis: By the way why is it relevant that Hu Jintao doesn't have the numbers in his closet somewhere? Because this kills the arguments of Gerald Segal and Gordon Chang. What they have argued is that Hu Jintao knows the real numbers, and so that whenever you see the Communist Party quote a number, its likely to be better than the real number which they know. So you take any number that is quoted in the official media, and the real situation is worse. I think Minxin Pei has done this from time to time. You see government official say that the total number of NPL's is $500 billion, well then, they must have secret statistics that say that the real number is $800 billion. What then happens is someone picks up $800 billion, and it becomes $1 trillion, and pretty soon China is on the verge of total economic and social collapse. By Twofish - 2/13/2008 4:48 PM

issac: the most worrisome part of national debt is the implicit -explicit pension , medical liability government assumed. Yes and no. It's worrisome because it is big and its going to get worse. It's not so worrisome because it's unlikely that you will wake up tomorrow or even six months from now, and have a sudden financial crisis that you didn't know about already. This is where capital structure makes a huge difference. US$200 billion in unfunded pensions is different than US$200 billion in loans based on bad real estate. The first situation you have five to ten years to figure out what to do. The second could turn into a big problem in weeks, days, or even hours. By Twofish - 2/13/2008 4:55 PM

I think Twofish has inadvertently proven the point Pettis is making. Estimates of the debt vary greatly. The less information and transparency there is, the greater the variation in the estimate. So far this must be very easy to understand. The additional point is that when the market is in a bull, the estimates are likely to be favorable to the borrower and when it is in a bear they are likley to be unfavorable. This adds volatility because it reinforces the market condition -- lack of transparency generally adds volatility for this reason. I think few experienced professional traders would disagree with this analysis. By ZH - 2/13/2008 9:13 PM

The Wall Street Journal quote is a good one. It shows that the effect of lack of transparency on volatility is not new. Have you read "The First Latin American Debt Crisis -- the City of London and the 1822-25 Loan Bubble" by Frank Dawson? It is a very good book and it makes the same point in an indirect way. By ZH - 2/13/2008 9:20 PM

Twofish has probably never been a trader or investor in a bad market which is why it may be difficult for him to understand how market dynamics affect the valuation of uncertainty. I would liike to hear more about the pension and other social welfare debt. Why isn't this a problem, Mr. Pettis, or if it is, why don't you discuss it in your article? By Moneybird - 2/13/2008 10:53 PM

ZH, thanks for your very elegant summary. Low transparency or poor information increases the range of estimates, and current market conditions tend to create systematic biases in our current estimates. These biases affect our behavior in a way that then exacerbates market conditions. I guess it is no big surprise, but in general markets with weak disclosure tend to be more volatile. Fatbrick said: "What do you mean on municipal and provincial obligations? Correct me if I am wrong, I think the municipal and provincial governments are not allowed to issue bonds in China." You are absoutely correct, Fatbrick, but they are allowed to borrow in other ways, usually from banks but also sometimes from other official entities. It is my understanding that all municipal and provincial borrowings, with the exception of certain project financings, are guaranteed by the central government, so on the bank books these are zero-risk loans, even if the borrowers are unable to pay, but they are nonetheless contingent liabilities of the central government, and are likey to become direct liabilities at exactly the wrong time. Isac and Moneybird, I agree with Twofish. The pension and social security obligations are, to paraphrase some 19th Century English diplomat, a disaster but not a problem. These are huge obligations and it is difficult to see how they will be met (by the way many other countries have the same problem), but they are far enough off in the future that they are not likely to have a major impact on near-term risk assesment. Twofish, it is not true that "Debt with guarantees that we don't know about isn't guaranteed debt." Putting wholly aside issues of explicit, implict and conditional guarantees, that is not even true in relatively open systems like those of the US and Europe and it certainly isn't true in China. To take a very simple case, there is a great deal of debate on whether the banks that were forced to hold dollars to bring down PBoC reserves had currency hedges with the PBoC. Many people assume that they do but none of us know for sure, nor do we know how these hedges are structured. Should we then assume that because we don't know, the hedges cannot exist? That seems a little foolish. The very fact of uncertainty will affect our perceptions of the value and existence of the hedges, and my experience tells me that our perceptions would mirror market sentiment at the time. Fatbrick, you are right that one difference between the US and China is that cedit risk is spread widely in the US whereas in China much of it ultimately resides in the central government. In my opinion that makes crises less likely but more severe, because as we have seen so many times in the past 200 years, when the sovereign credit is questioned the country can suffer from very severe financial distress. That is why I think it is so important to get a clear understanding of the sovereign balance sheet and for the government to implement measures that reduce its probability of default. For example, the central government should not guarantee provincial and municipal debt -- it can be easily proved that the "savings" so obtained exist only for specific transactions and are non-existent at the macro-level, and while those guarantees reduce the probability of small, localized crises, they increase the probability of a large, macro crisis. By Michael Pettis - 2/14/2008 4:39 AM

Well. the crisis had happened in China in 1990s. The real estate bubble bursted around 1993. There was a high inflation. Some regional banks became insolvent. I remembered that the creditors then were some Japanese banks. They were quite shock when they heard that Chinese central government would not back the troubled regional banks and state-owned trust funds. Back then there was no WTO for China. Local officials had much more say on bank loan. Banks had higher bad loan ratio and much less capital. The real estate and stock markets were crashed... The upside now is that it is pretty unlikely that we are going to that level again. By fatbrick - 2/14/2008 5:38 AM

Pettis: It is my understanding that all municipal and provincial borrowings, with the exception of certain project financings, are guaranteed by the central government, so on the bank books these are zero-risk loans, even if the borrowers are unable to pay, but they are nonetheless contingent liabilities of the central government, and are likey to become direct liabilities at exactly the wrong time. This is very incorrect, which may account for the fact that I think that I think that the Central Government's liabilties are not as large as you think they are. Provincial and local governments must by law maintain balanced budgets, and have absolutely no authority to borrow money directly. They do borrow money by having companies controlled by the provincial or local government borrow money and then have that money transferred to the provincial or local government. This borrowing is first of all, not the debt of the local or provincial government but rather the debt of the company controlled by the local or provincial government and hence the company can default without legal liability. The second thing is that this sort of debt is totally illegal. It is against the budget law for a local or provincial government to operate a company to funnel money to the local operating budget, and so if the central government wants to walk away from this debt, it can declare the loans illegal (which they are) and not pay a cent. Now the central government *may* step and pay the debts anyway, if it affected the solvency of a major bank or would cause mass unemployment, but a lot of loans are made through local controlled trust and investment companies or credit cooperatives. If those went bad, then the central government could and would walk away from those debts. This happened in 1998 with the Guangdong International Trust Investment Corporation, in which the government refused to bail out the company, and investors got maybe 10 cents/dollar. By Twofish - 2/14/2008 10:19 AM

Pettis: For example, the central government should not guarantee provincial and municipal debt They don't, and they have made it very clear that they don't. This is one of the lessons that the Chinese government learned from Latin America, when you had this huge problem in places like Brazil, where subnational entities could borrow money with the expectation that the national government would bail them out. This naturally led to a huge mess. By Twofish - 2/14/2008 10:23 AM

This is dragging out a little, but there is clearly a lot of provincial and municipal debt in China. We should not confuse the issuance of bonds with the existence of debt. Here is Liu Ligang of REITI has to say about provincial and municipal debt in a presentation about reforming the municipal and provincial debt system: Local government debt problem is quite serious •    Five types of local government debt •    Guarantees via the “decisions” issued by local NPCs on loans from banks to local government-owned enterprises (explicit contingent debt) •    Using social security fund for current expenditure needs (implicit but direct debt) •    Losses due to state-owned grain related enterprises or grain debt (implicit but direct debt) •    Salary non-payment to…government employees (explicit debt) •    Construction debt owned to construction firms (about one percent of GDP By Michael Pettis - 2/14/2008 11:41 AM

That's a pretty good categorization of the debt problem. 1) is interesting because the local people's congresses do not have legal authority to guarantee payment of debt of local government-owned enterprises from state funds. It's unlikely that the big state banks would take such a guarantee, Trouble is that if you have a local bank or quasi-bank that is owned by the same government that is trying to raise money...... The main way of reforming the debt system is to fund provincial and local governments directly with central government funds so that they don't have to go into debt, which is what Beijing has been doing over the last three years. It's really important in looking that these things to look at the details rather the total aggregate numbers. Each piece of debt has its own characteristics, and you can easily have a crisis in what seems to be a small liability explodes in response to some crisis. The other thing can happen in which you have a massive huge debt that *doesn't* explode because it is structured in a way that doesn't make it due all at once. Pensions are an example of this. It's also important to disaggregate debt for public policy reasons. Giving a big number of debt, and then saying "reduce this debt" is not useful since you end up reducing the wrong debt and making bad social choices. If you have a detailed breakdown of the debt, you can go through and point by point do things to make sure that things go under control, and something that is a counterpart to the list of local debt is what the government has been doing to keep things under control. By Twofish - 2/14/2008 1:50 PM

Also, the central government can get rid of pretty much all of its salary debt to government employees by asking them just exactly how they were able to survive without the government paying their salaries. The deal would be that the employee doesn't raise the issue of back pay if the government doesn't raise the issue of what the employee did to substitute for the lack of payment. One has to ask why one would want to be a government official if one's salary isn't being paid, and the answer is that one is receiving a salary, just not one that one would want to talk loudly about. By Twofish - 2/14/2008 8:26 PM