Excerpted from Michael Pettis latest letter to investors,
Clearly in the past two or three years there has been a huge shift in market perceptions of Chinese debt. Everyone recognizes that debt has become a serious problem.
But there are two very different ways to recognize this. Some believe that the Chinese financial system, and perhaps the shadow banking system more specifically, took a number of wrong steps, compounded by the lack of discipline among local governments, and created a debt problem. By that logic, Beijing can take administrative steps to address credit creation and to bring debt under control. It is especially important, according to this view, to analyze the source of risky credit creation and to suppress it, which is perhaps why so much attention of late has been placed on the shadow banking system and on ways in which Beijing can “resolve” the existing stock of bad debt.
My view is different. Burgeoning debt was not an unlucky accident. It is fundamental to the way the growth model works, and we have arrived at the stage, probably described most imaginatively by Hyman Minsky in his work on balance sheets, in which the system requires an acceleration in credit growth simply to maintain existing levels of economic activity.
China’s debt problems, in other words, cannot be resolved administratively, by fixing the shadow banking system, by imposing discipline on borrowers, or indeed by eliminating financial repression (much of which, by the way, has already been squeezed out of the system by lower nominal GDP growth). Without a massive transfer of wealth from the state sector to the household sector it will be impossible, I would argue, for GDP growth rates of anything above 3-4% – and perhaps even less – to occur without a further unsustainable increase in debt, whether that increase occurs inside or outside the formal banking system and whether or not discipline has been imposed on borrowers.
Last month I spoke with a very prominent European economist and he assured me that although he now agrees (he used strongly to deny it) that China has debt “problem”, he believes it can easily be resolved by “socializing” the debt, by which he means transferring it onto the government balance sheet.
I disagree completely, and not just because transferring bad debt from local governments to the central government, while undoubtedly reducing the probability of a legal default, does not in the slightest way address the cost of resolving the bad debt. The “successful” previous bailouts were not successful in any way if you place their “success” in the context of the rebalancing process, and this is obvious if you work through the full consequences on the structure of Chinese demand.
Because economists for many years have been trained to ignore balance sheets and, more generally, the way debt drives economic activity, the quality of analysis, especially the analysis of economic turning points during which the amount and structure of debt can create significant constraints on the way rebalancing can occur, has devolved, to be replaced mostly by vaguely empirical and very mathematically confused constructions.
I have described many times before why excess credit creation is at the heart of much of China’s GDP growth, and why this means that China must choose between a sharp slowdown in GDP growth as credit is constrained, or a continued unsustainable increase in debt.
No other options are available. But even this point is about new credit creation does not address the existing stock of bad debt, which is what I want to discuss in this blog entry. If you assume that for many years China has been misallocating investment (by which I simply mean that the resulting increase in productivity generated by the investment was less than the correctly calculated debt-servicing cost), it should be obvious that because there have been almost no defaults or other forms of debt write-down, the implicit losses have simply been rolled over, most likely in the balance sheets of the Chinese banks. This has several implications:
1. GDP growth has been implicitly increased by the amount of losses that should have been, but were not, written down. This means that China’s GDP today, compared to countries in which it is more difficult simply to roll over losses indefinitely, is overstated, and I suspect that it may be overstated by as much as 20-30%. Why? Because in an economy in which losses were not simply accumulated and rolled over, the amount of the write-down (which would have occurred, either as a default, or as an equivalent transfer from a more profitable part of operations to subsidize the loss) would have shown up as lower GDP.
2. In that case all GDP-related data is biased in a predictable way. Productivity numbers, for example, are biased upwards, and real worker’s productivity is lower than the numbers posted officially.
3. Losses that are rolled over do not disappear. They are implicitly amortized over the period of the loan, which, assuming that loans are rolled over indefinitely, means that every year a declining portion of that loan is effectively written down. Over long periods of time every economy recognizes investment losses, but depending on how these losses are treated, the recognition can take place either in the period in which the losses occur or over the loan amortization period.
4. There is a lot of confusion over how the implicit amortization of unrecognized losses takes place over time. Let us assume that an investor borrows $100 to invest in a project that creates only $80 of value. The project, in other words, creates a loss of $20. If the loss is not immediately recognized, there is a gap between the true economic value of the debt servicing cost and the increase in productivity associated with the project. This gap must be covered by implicit transfers from some other part of the economy, and these transfers reduce the economic activity that would have otherwise been created.If the gap is covered by financial repression, for example, (i.e. the authorities force down the borrowing cost to less than the increase in productivity generated by the project, so that the borrow shows a profit), the cost of amortizing the loss is passed onto the net lenders (usually, but not always, the household sector, who are net lenders to the banking system) in the form of a lower return on their savings. This lower return reduces their total income and, in so doing, reduces their consumption, which effectively reduces future GDP growth by reducing demand.
5. GDP growth is only artificially boosted during the period in which the total amount of losses rolled over exceeds the amount of the amortization. After that GDP growth is artificially constrained. When the system is still accumulating and rolling over losses, in other words, GDP growth is systematically biased upwards. When it stops, GDP growth will be systematically biased downwards.
6. This bias can be considerable. Let us assume, for example, that the real growth in an economy causes it to double its wealth every 10 years. Real GDP would, in that case, increase every year on average by nearly 7.2%. Let us also assume that during the first ten years, GDP growth was overstated by a failure to recognize investment losses, so that reported GDP growth was actually 10%. Finally we will assume that after ten years, this over-reporting stopped, and the excess GDP was amortized during the next ten years so that at the end of twenty years GDP was once more correctly stated. The numbers how that at the end of ten years, reported GDP would be overstated by 22.9% – that is, instead of doubling, reported GDP would be 159% higher. During the next ten years, as real GDP continued to grow by 7.2%, reported GDP would grow on average by just over 4.4% as the earlier losses that had not been recognized were amortized.
7. My numbers above assume that the overstatement and understatement are symmetrical. In fact the process is not symmetrical because of the possibility of financial distress costs. The total value of overstated GDP during the period when losses are being rolled over is only equal to the total value of the subsequent amortization of those losses if there are no financial distress costs.
8. But there are in fact likely to be substantial financial distress costs. In corporate finance theory we have a very clear understanding of how high debt levels change incentive structures in such a way so as to reduce overall growth. This means that the longer it takes to amortize the hidden losses, the greater the amount by which the future amortization costs will exceed the current overstatement of GDP. Japan after 1990 might a good example of this process. Its share of global GDP rose from roughly 10% in 1980 to 17% at is peak, only to have declined since then to roughly 9% of global GDP. This is an astonishing relative decline, and it must have been made worse at least in part by the financial distress costs imposed on an economy unwilling to write down debt correctly.
9. Remember that the only way debt can be resolved is by assigning the losses, either during the period in which the losses occurred or during the subsequent amortization period. There is no other way to “resolve” bad debt – the loss must be assigned, today or tomorrow, to some sector of the economy. “Socializing” the debt, or transferring the debt from one entity to another, does not change this.
10. There are three sectors to whom the cost can be assigned: households, businesses, or the government. In China we might usefully think of these as households, small and medium enterprises (SMEs), and the state sector (in principle there is a fourth sector, foreigners, to whom the losses can be assigned, but it is very unlikely that they will bear much of the losses). It is pretty clear that after the banking crisis of the late 1990s, the losses were assigned, largely in the form of financial repression, to the household sector.
11. To the extent that China has significant hidden losses embedded in the balance sheets of the banks and the shadow banks, over the next several years Beijing must decide how to assign the losses. If it assigns them to the household sector, it will put significant downward pressure both on household income growth (which will be less than GDP growth) and, consequently, on consumption growth. Rebalancing means effectively that consumption growth (and household income growth) must exceed GDP growth, which means that even if GDP growth slows to 3-4%, as I expect, household income can continue growing at 5-6%. This explains why, contrary to the consensus, a more slowly growing, rebalancing China will not lead to social unrest.
12. Of course if the losses are assigned to the household sector, China cannot rebalance and it will be more than ever dependent on investment to drive growth. This is why I reject absolutely the argument that because China resolved the last banking crisis “painlessly”, it can do so again.
13. Beijing can also assign the losses to SMEs. In effect this is what it started to do in 2010-11 when wages rose sharply (SMEs tend to be labor intensive). It is widely recognized that SMEs are the most efficient part of the Chinese economy, however, and that assigning the losses to them will undermine the engine of China’s future productivity growth.
14. Finally Beijing can assign the losses to the state sector, by reforming the houkou system, land reform, interest rate and currency reform, financial sector governance reform, privatization, etc. Most of the Third Plenum reforms are simply ways of assigning the cost of rebalancing, which includes the recognition of earlier losses, to the state sector. This is likely however to be politically difficult. China’s elite generally benefits tremendously from control of state sector assets, and they are likely to resist strongly any attempt to assign to them the losses.
This is how I think we need to think about China’s debt problem. Notice that I am making no predictions. I am only trying to outline as schematically as possible the only ways in which the debt problem can be resolved. There are no other possible ways to address the debt, and so any analysis we do or propose must be consistent with the model described above.
The key point is that we cannot simply put the bad debt behind us once the economy is “reformed” and project growth as if nothing happened. Earlier losses are still unrecognized and hidden in the country’s various balance sheets. These losses will either be explicitly recognized or they will be implicitly amortized. The only interesting question, as I see it, is which sector will effectively be assigned the losses. This is a political question above all, and its answer will tell us a great deal about how the newly-constituted, “reformed” China will grow over the next few decades.