全球智库信息集成服务系统

A recurring conversation I have with clients concerns the ability of banks to create credit, and of governments to monetize debt, and whether this ability is the solution to or the cause of financial instability and economic crisis. Monetarists and structuralists (to use Michael Hudson’s names for the two sides, whose centuries-long debate, exemplified by Thomas Malthus and David Ricardo during the Bullionist Controversy, dominates the history of economic thinking) have very different answers to that question, but I will suggest that each side disagrees because it implicitly assumes an idealized version of an economy.

We are normally taught that banks allocate credit by lending the money that savers have deposited in the banking system, but in fact banks create deposits in the banking system by creating credit, so it seems to many as if they can create demand out of nothing. Similarly, if governments are able to create money, and if they can borrow in their own currency, they can easily monetize debt, seemingly at no cost, by “printing” the money they need to repay the debt (actually by crediting bank accounts, which amounts to the same thing). This means that when they borrow, rather than repay by raising taxes in the future, all they have to do is monetize the debt by printing the money needed to repay the debt. It seems that governments too can create demand out of nothing, simply by deficit spending.

There is a rising consensus – correct, I think – that the misuse of these two processes – which together are, I think, what we mean by “endogenous money” – were at the heart of the debt surge that was mischaracterized as “the great Moderation”. For example in a book published earlier this month,Between Debt and the Devil, in which he provides a description of the rise of debt financing in the four decades before the 2008-09 crisis, along with the economic risks that this has created, Adair Turner specifies these two as fundamental to the rising role of finance in the global economy. He writes:

…in modern economics we have essentially two ways to produce permanent increases in nominal demand: either government fiat money creation or private credit money creation.

I am less than half-way through this very interesting book, so I am not sure how he addresses the main characteristics of debt, nor whether he is able to explain how much debt is excessive, or identify the main ways in which the liability side of the macroeconomic balance sheet intermediates behavior on the asset side to determine the growth and volatility of an economy. He invokes the work of Hyman Minsky often enough, however, to suggest that unlike traditional economists he fully recognizes the importance of debt.

And it is because of this importance that the tremendous confusion about what it means to create demand out of nothing is dangerous. When banks or governments create demand “out of this air”, either by creating bank loans, or by deficit spending, they are always doing one or some combination of two things, as I will show. In some easily specified cases they are simply transferring demand from one sector of the economy to themselves. In other equally easily specified cases they are creating demand for goods and services by simultaneously creating the production of those goods and services. They never simply create demand “out of thin air”, as many analysts seem to think, and doing so would violate the basic accounting identity that equates total savings in a closed system with total investment.

I had originally intended this blog entry to be a response to questions in some of the comments following my last blog post, but because my response turned out to be too long to submit as a comment, and because the questions lead to a far more complex answer than I had originally planned, it has become a blog entry in its own right. The questions arise in the context of a discussion of some of Steve Keen’s work among several regular commenters on my blog. Keen is an Australian post-Keynesian who heads the School of Economics, History and Politics at Kingston University in London.

I’ve known of Keen’s work for many years, and last year he spoke at my PKU seminar on central banking (as has Adair Turner, by the way). He is one of the most hard-core proponents of Hyman Minsky, and regular readers know that I think of Minsky as one of the greatest economists since Keynes. In the third chapter of my 2001 book, The Volatility Machine, I explain the ways in which developing countries designed balance sheets that systematically exacerbated volatility – and which eventually led to debt-based contractions or financial crises – in terms of a framework that emerges from the work of Minsky and Charles Kindleberger. This framework – something that many Latin American economists have no trouble understanding but which has been ignored by nearly all Chinese and foreign economists covering China – explains why three decades of economic expansion in China, underpinned by rapid growth in credit and investment, would lead almost inevitably to destabilizing debt structures.

Hyman Minsky’s balance sheets

Minsky is important not so much for the “Minsky Moment”, a phrase he never used, but rather because of his profoundly intuitive balance-sheet oriented understanding of the economy, something that set him completely apart from most contemporary academic economists who, for the most part, have barely begun to incorporate balance sheet dynamics into their analyses. Minsky’s insights include his now-well-known description of accelerating financial fragility, along with his explanation of why instability is inherent to the financial sector in a capitalist economy.

Most insightful of all, Minsky characterized the economy as a system of interlocking balance sheets, and because he taught us to think of every economic entity as effectively a kind of bank, with one entity’s assets being another’s liabilities, it follows that economic performance is partly a function of the direction and the extent in which the two sides of each balance sheet are mismatched. Because these mismatches vary as a consequence of past conditions and future expectations, when institutional distortions are deep, balance sheet mismatches in the aggregate can be systemic, in which case they determine how an economic system behaves and responds to exogenous and endogenous shocks.

Minsky’s framework made it especially easy to predict the difficulties that China would face once it began to rebalance its economy. China can be described as an extremely muscular illustration of Minsky’s famous dictum that “stability is destabilizing”. Its financial system was designed to meet China’s early need for rapid credit expansion, and it evolved around what seemed like permanently high growth rates and uninterrupted access to financing. Two decades of “miracle” levels of investment-driven growth, the role of the financial sector in that growth, and the unrealistic expectations that Chinese businesses, banks, and government entities had consequently developed, reinforced by sell-side cheerleaders, made it obvious that the interlocking balance sheets that make up the Chinese economy had added what was effectively a highly “speculative” structure onto the way economic entities financed their operations.

This would sharply enhance growth rates during the expansion phase, much like margin borrowing enhances returns when market prices are rising faster than the debt servicing costs, but at the expense of sub-par performance once conditions reverse. The process is actually quite easy to describe, and the fact that it caught nearly the entire community of analysts by surprise should indicate just how unfamiliar economists are with the approach championed by Minsky.

Ignoring the balance sheet framework does not always result in bad economics. When debt levels are low, and the economy close to a kind of Adam Smith type of economy, in which there are no institutional constraints and no entities large or important enough to affect the system as a whole, it makes sense to ignore liabilities and to analyze an economy only from the asset side in order to understand and forecast growth. Evaluating only the asset side would still be conceptually wrong, because both sides of the balance sheet always matter, but the difference between analyses that ignore the liability side and analyses that incorporate the liability side are small enough to ignore.

When conditions change in certain ways, however, the differences can become too large to ignore. The more deeply unbalanced an economy, the higher its debt levels, or the more highly systematically distorted its balance sheets, the more the two forecasts will diverge and the more urgent it is that economists incorporate the balance sheet in their analyses.

In a way it is like an engineer who builds a bridge using Newton’s equations rather than Einstein’s. In a motionless world, or in the close approximation in which most of us live, Newtonian errors are insignificant, and the bridge the engineer builds will carry traffic almost exactly as expected. As objects accelerate, however, these small errors eventually become vast, and the Newtonian bridge risks becoming useless.

In the early 1990s the models that most economists used to analyze and explain Chinese economic growth were good enough, like the Newtonian bridge in the slow moving world in which humans operate. By the late 1990s, however, the sheer extent of bad debt within the banking system should have provided a warning that mismatches and imbalances might have become large enough to invalidate the old models. They clearly did invalidate the old models over the next few years as credit misallocation accelerated, along with the depth and direction of now-unprecedented imbalances and highly self-reinforcing price changes in commodities, real estate, stock markets, and other variables – what George Soros might have cited as extreme cases of reflexivity.

Violating identities

To get back to the discussion in the comments section, a very brisk and active debate broke out among a number of readers over Keen’s claim that next period growth is a function of both this period’s economic conditions as well as this period’s change in debt. I won’t summarize the discussion, which is long and wide ranging, but part of the disagreements have to do with whether Keen’s dynamic model, which incorporates changes in debt, implies that the accounting identities I use are somehow invalid.

One reader, Vinezi, wrote “Michael has been repeatedly saying that he is using the same identities as the basis of his research for the last 10 years. All his insights presented on this blog, which, in my opinion, are spectacularly correct, are derived from the application of these very identities”. Vinezi then goes on to ask for my response to Keen’s rejection of these identities, most importantly the identity between savings and investment.

I don’t know if Keen actually rejects the identity, but I doubt that he does because he is too good a mathematician not to know that identities cannot be “accepted” or “rejected” like hypotheses or models. More generally I would never say that I am using this (or any other) identity as the basis for my research, as Vinezi states, because the point of research is, or should be, to test hypotheses (or, in a common but sloppy practice, to discover hypotheses). You cannot “test” accounting identities, however, because they are not hypothetical. They are true either by definition or as a logical necessity (which may well be the same thing), and there is no chance that they can be wrong.

I do refer often to basic accounting identities, but mainly because too many economists and analysts allow themselves to become so confused by balance of payments arithmetic, money creation, and so on, that they try to explain the relationships among different variables by proposing hypotheses that violate accounting identities. In that case their hypotheses are simply wrong, and rejecting them does not require any empirical support. Rather than use empirical data to “test” the identities, it is more accurate to use the accounting identities to “test’ the data. If the data seems to violate the identities, then it must be the case that the data is incorrectly collected or incorrectly interpreted.

The important point about accounting identities – and this is so obvious to logical thinkers that they usually do not realize how little most people, even extremely intelligent and knowledgeable people, understand why it matters – is that they do not prove anything, nor do they create any knowledge or insight. Instead they frame reality by limiting the number of logically possible hypotheses. Statements that violate the identities are self-contradictory and can be safely rejected.

Accounting identities are useful, in other words, in the same way that logic or arithmetic is useful. The relevant identities make it easier to recognize and identify assumptions that are explicitly or implicitly part of any model, and this is a far more useful quality than it might at first seem. Aside from false precision, my biggest criticism of the way economists use complex math models is that they too-often fail to identify the assumptions implicit in the models they are using, probably because they are confused by the math, and they would often be forced to do so if they weren’t so quick dismiss accounting identities on the grounds that these identities don’t tell you anything about the economy.

It’s true, they don’t. But arithmetic doesn’t tell you anything about how to build a bridge either. Unlike economists, however, engineers have no choice but to stick rigidly to the identities. While economists tolerate models that are not constrained by accounting identities because, for some reason, economists do not seem constrained by the need for their models of the economy to conform to reality, any engineer whose model for a bridge requires that two plus three equal seven would find it hard to build bridges, harder to find clients, and even harder to get a teaching job at any university. Remembering always to maintain accounting identities does not lead to true statements or to brilliant insights, but it does make it easy to reject a very large class of false or muddled statements. Just as logic doesn’t create science, but it prevents us from making bad science, identities do not create models, but they protect us from useless models.

Keynes, who besides being one of the most intelligent people of the 20thcentury was also so ferociously logical (and these two qualities do not necessarily overlap) that he was almost certainly incapable of making a logical mistake or of forgetting accounting identities. Not everyone appreciated his logic. For example his also-brilliant contemporary (but perhaps less than absolutely logical), Ralph Hawtrey, was “sharply critical of Keynes’s tendency to argue from definitions rather than from causal relationships”, according to FTC economist David Glasner, whose gem of a blog, Uneasy Money, is dedicated to reviving interest in the work of Ralph Hawtrey. In a recent entry Glasner quotes Hawtrey:

[A]n essential step in [Keynes’s] train of reasoning is the proposition that investment an saving are necessarily equal. That proposition Mr. Keynes never really establishes; he evades the necessity doing so by defining investment and saving as different names for the same thing. He so defines income to be the same thing as output, and therefore, if investment is the excess of output over consumption, and saving is the excess of income over consumption, the two are identical. Identity so established cannot prove anything. The idea that a tendency for investment and saving to become different has to be counteracted by an expansion or contraction of the total of incomes is an absurdity; such a tendency cannot strain the economic system, it can only strain Mr. Keynes’s vocabulary.

This is a very typical criticism of certain kinds of logical thinking in economics, and of course it misses the point because Keynes is not arguing from definition. It is certainly true that “identity so established cannot prove anything”, if by that we mean creating or supporting a hypothesis, but Keynes does not use identities to prove any creation. He uses them for at least two reasons. First, because accounting identities cannot be violated, any model or hypothesis whose logical corollaries or conclusions implicitly violate an accounting identity is automatically wrong, and the model can be safely ignored. Second, and much more usefully, even when accounting identities have not been explicitly violated, by identifying the relevant identities we can make explicit the sometimes very fuzzy assumptions that are implicit to the model an analyst is using, and focus the discussion, appropriately, on these assumptions.

No surpluses on both the capital and current accounts

A case in point is The Economic Consequences of the Peace, the heart of whose argument rests on one of those accounting identities that are both obvious and easily ignored. When Keynes wrote the book, several members of the Entente – dominated by England, France, and the United States – were determined to force Germany to make reparations payments that were extraordinarily high relative to the economy’s productive capacity. They also demanded, especially France, conditions that would protect them from Germany’s export prowess (including the expropriation of coal mines, trains, rails, and capital equipment) while they rebuilt their shattered manufacturing capacity and infrastructure.

The argument Keynes made in objecting to these policies demands was based on a very simple accounting identity, namely that the balance of payments for any country must balance, i.e. it must always add to zero. The various demands made by France, Belgium, England and the other countries that had been ravaged by war were mutually contradictory when expressed in balance of payments terms, and if this wasn’t obvious to the former belligerents, it should be once they were reminded of the identity that required outflows to be perfectly matched by inflows.

If Germany had to make substantial reparation payments, Keynes explained, Germany’s capital account would tend towards a massive deficit. The accounting identity made clear that there were only three possible ways that together could resolve the capital account imbalance. First, Germany could draw down against its gold supply, liquidate its foreign assets, and sell domestic assets to foreigners, including art, real estate, and factories. The problem here was that Germany simply did not have anywhere near enough gold or transferable assets left after it had paid for the war, and it was hard to imagine any sustainable way of liquidating real estate. This option was always a non-starter.

Second, Germany could run massive current account surpluses to match the reparations payments. The obvious problem here, of course, was that this was unacceptable to the belligerents, especially France, because it meant that German manufacturing would displace their own, both at home and among their export clients. Finally, Germany could borrow every year an amount equal to its annual capital and current account deficits. For a few years during the heyday of the 1920s bubble, Germany was able to do just this, borrowing more than half of its reparation payments from the US markets, but much of this borrowing occurred because the great hyperinflation of the early 1920s had wiped out the country’s debt burden. But as German debt grew once again after the hyperinflation, so did the reluctance to continue to fund reparations payments. It should have been obvious anyway that American banks would never accept funding the full amount of the reparations bill.

What the Entente wanted, in other words, required an unrealistic resolution of the need to balance inflows and outflows. Keynes resorted to accounting identities not to generate a model of reparations, but rather to show that the existing model implicit in the negotiations was contradictory. The identity should have made it clear that because of assumptions about what Germany could and couldn’t do, the global economy in the 1920s was being built around a set of imbalances whose smooth resolution required a set of circumstances that were either logically inconsistent or unsustainable. For that reason they would necessarily be resolved in a very disruptive way, one that required out of arithmetical necessity a substantial number of sovereign defaults. Of course this is what happened.

The same kind of exercise eight-five years later, shortly after the euro crisis, made it clear that Europe was limited by similar accounting identities to three options. First, Germany could reflate domestic demand by enough to exceed the consequent increase in its domestic production of goods and services by at least 4-5% of GDP, and probably more (i.e. it had to run a current account deficit). Second, peripheral Europe could tolerate excruciatingly high unemployment for at least a decade, and probably more.

Third, peripheral Europe could leave the euro and restructure its debt with substantial debt forgiveness – or, which is nearly the same, force Germany to leave the euro, which would require much less debt forgiveness – causing losses in the German banking system at the same time that it caused Germany’s manufacturing sector to drop precipitously. (A fourth option, that Europe could run huge surpluses with the rest of the world, perhaps two times or more than its current surplus, was too implausible to consider, and although Europe is certainly running irresponsibly high surpluses, they are not high enough to allow Europe to grow.) So far Europe has chosen the second option, with a high probability, in my opinion, that before the end of the decade it will be forced into the third.

This is why we must keep accounting identities firmly in mind. They don’t tell us what to do, but keeping them in mind prevents us from proposing, or believing arguments, that are clearly inconsistent, or often simply idiotic. To take another immediate example, one of the few recent bits of cheer in our otherwise very glum world has been the almost teenagerish excitement with which David Cameron has been BFFing. It’s not all just unconditional friendship, however, and apparently he hopes to get big deals and significant inward investment announced in the next week. It sounds good, but, a firm grasp on the accounting identities would identify which kinds of “big deals” are likely to boost GDP and which merely to shift the locus of GDP creation.

More importantly, it would show that for a rich, developed country like England, inward investment almost always affects growth adversely (unless it brings technological and managerial advances with it) and never more obviously so than when interest rates are struggling against the zero bound and every country is urgently trying to export excess savings. As one of my exasperated PKU students asked me after class last Saturday when we discussed the president’s trip: “So everyone agrees that it is good for England to get much more foreign investment, and everyone also agrees that it is bad for England to have a much bigger trade deficit. Don’t they know it’s the same thing?”

Not everyone does, but to return to the reference I made to discussion in the comments section that started this essay, regardless of whether or not Vinezi is correctly interpreting Steve Keen on the savings and investment identity, does his claim – that an increase in debt causes a corresponding increase in GDP growth (and the conditions under which this is likely to be true correspond closely, I think, with current conditions) – imply that investment can exceed savings? Or as Vinezi puts it:

Steve Keen, ErikWim, Suvy & Willy2 claim that the mistake Michael makes is that he is using a “loanable funds model” in which savings and investment are “merely being matched with each other”. Steve Keen, ErikWim, Suvy & Willy2 are pointing to the new “endogenous money model” of the modern-banking sector in which investments can be made even without having the savings a priori. Yes? Would the “Steve Keenites” here please confirm that this is how all of you would like to correct Michael’s “flawed” identity? Michael, if you read this, would you please respond to their attack on your most fundamental research assumption?

Before responding I have two parenthetical responses. First, the savings and investment identity is not my “most fundamental research assumption” because it is not an assumption at all, and it cannot be meaningfully used in research. Second, the loanable funds model does indeed permit credit to be created “out of thin air” (it is perhaps what we would call the neoclassical tradition that doesn’t, although I have to admit I don’t always keep the lines between different traditions terribly sharp), and while there is much that I find deeply insightful in both Knut Wicksell – and I assume that his “cumulative process” is part of the intellectual tradition on which “loanable funds” is based – and his rival Irving Fisher, I don’t think of the work of either of them as supporting or opposing my use of the saving and investment identity to understand global imbalances.

It might seem that Wicksell denies the savings and investment identity because he says that the two are equal only when the money interest rate is set equal to the natural interest rate. We see that in a 1986 paper by Richmond Fed economist Thomas Humphrey (whose books are unfortunately out of print):

The cumulative process analysis itself attributes monetary and price level changes to discrepancies between two interest rates. One, the market or money rate, is the rate that banks charge on loans. The other is the natural or equilibrium rate that equates real saving with investment at full employment and that also corresponds to the marginal productivity of capital. When the loan rate falls below the natural rate, investors demand more funds from the banking system than are deposited there by savers. Assuming banks accommodate these extra loan demands by issuing more notes and creating more demand deposits, a monetary expansion occurs. This expansion, by underwriting the excess demand for goods generated by the gap between investment and saving, leads to a persistent and cumulative rise in prices for as long as the interest differential lasts.

This might seem indeed to violate my claim that any model that requires or even permits global investment to exceed savings is logically impossible, but this is only because the difference lies in what economists call the ex antequantities. This means that at any given money interest rate (other than the natural interest rate), desired savings may differ from desired investment, but one or the other (or both) must adjust so that in the end they do equate, the result being a sub-optimal amount of investment. Excessively low interest rates in China (until 2012) meant, for example, that desired investment was far too high, and much higher than desired savings, but in a financially repressed system, as I have shown many times, low interest rates can actually force up savings by constraining the household income share of GDP, which is what happened not just in China before 2012 but also in Japan in the 1980s.

I think what Keen might actually be saying is that if investment in the next period is greater than savings in the current period – if it is boosted, so the argument goes, by the ability of the banking system to fund investment by creating debt “out of thin air” – this does not violate the identity, and it is not only possible, but even likely. (If by any chance Steve keen should read this, perhaps he might respond.)

Creating demand “out of thin air”

But it is possible not because banks can fund investment by creating debt “out of thin air”. This statement is either highly confused or it too-easily leads others into confusion. There is a related form of this question that often seems to come out of the MMT framework, although I have no idea if this is a misreading of MMT or if it is fundamental to the theory, but while banks can create debt, they do not automatically create additional demand. According to MMT, as I understand it, there is no limit to fiscal deficits because governments who control the creation of money can repay all obligations regardless of their taxing capacity simply by monetizing the debt (which of course means nothing more than exchanging debt which we call “debt” for debt which we call “money”).

A lot of people seem to think that this Because savings and investment must always balance, the idea that the savings rate in any country is determined at home is nonsense.