The right way to create a good bank and a bad bank

February 23, 2009

Policymakers continue to struggle to figure out how to turn a troubled bank into a good bank and a bad bank. Under the good-bank/bad-bank policy, the good bank will operate free from concerns about troubled assets, because these assets will be held by the fully independent bad bank. Most discussions of the separation of a bank in this way presume that the government must inject a lot of new capital to create a well-capitalized good bank together with a still-solvent bad bank. The math seems simple–the troubled bank has almost no capital, so if the capital has to be split between the two banks, a well-capitalized bank will need new capital.

But this math is incomplete. We can create a good bank with a big cushion of capital while keeping the bad bank as solvent as the existing integrated bank. The key idea–from Jeremy Bulow–is that the bad bank owns all of the equity in the good bank.

We illustrate with data from Citicorp (in billions of dollars):

citiThe left column shows Citicorp’s balance sheet roughly marked to market. The company’s value in the stock market of $11 billion is $76 billion less than reported book equity value. We deduct that amount from the reported value of long-term assets, which is where the troubled real-estate related assets are most likely to reside.

The other two columns show the balance sheets of the new good bank and bad bank. The good bank will continue to operate under the Citi brands as a well-capitalized operating entity. The bad bank will be a financial fund with no operating functions.  The good bank gets the short-terms assets and the “other” assets because many of these are related to its operating activities. It gets the better half of the long-term assets, taken to have book value, while the bad bank gets the poor half, where the impairment has already occurred and suspicions of further price declines persist. The bad bank holds the valuable equity in the good bank to the tune of $427 billion.

The deposits remain as liabilities of the good bank. Because the good bank is heavily capitalized, the deposits are safe. Most are uninsured, so the creation of the good bank eliminates the danger of a run on the bank by those depositors. All of the debt goes to the bad bank. The holders of the debt were never promised a government guarantee. The shareholders in Citicorp become the shareholders in the bad bank. They are indirectly shareholders in the good bank as well, because the bad bank owns the good bank.

The bad bank is thinly capitalized. In fact, it has exactly the same amount of capital that Citi had in the first place. With further declines in the values of the troubled assets, the bad bank may become insolvent. In that case, the bondholders will need to negotiate diminished values or the bad bank will need to be reorganized. In either case, the shareholders will lose all their value, just as they would have lost that value had Citi not been divided and there had been no further bailout from the government. The bondholders will lose part of their value, because there is no reason or justification for bailing them out.

The key property of this approach is that neither the shareholders nor the bondholders suffer any impairment of their existing intrinsic values. At present, the shareholders own an out-of-the-money call option on the assets. The bondholders own the assets subject to the shareholders’ call. Their combined value declines by the amount of any further decline in the value of the troubled assets. As claimants on the bad bank, they would be in the same situation, sharing the same decline in value, with the losses of bad-bank shareholders equal to the losses of Citi’s shareholders absent the separation and the same for the bondholders.

The adoption of the proposed good- and bad-bank separation would result in capital losses for the shareholders and bondholders, because the new policy would eliminate the benefit that they might receive from further bailout money from the government. The potential reorganization of the bad bank made necessary by future insolvency would not create any kind of financial emergency, so there would be no reason for the government to bail out the bad bank. There is no reason not to inflict the capital losses on the shareholders and bondholders, as they represent the capitalization of possible bad policies and are unrelated to the assets that the shareholders and bondholders actually own.

Now one might wonder, given that the intrinsic claims of neither the bond holders nor the equity holders has changed, why bother?  Pundits talk about the toxic assets in the banking system as if somehow they were infectious, and the good assets would become infected by the bad assets. One envisions the mold on one piece of cheese taking up residence on an adjacent piece in the frig. Or that somehow if the bad, hard-to-value, assets were moved somewhere else, both sets of assets would be easier to value and the banking system somehow more sound.  We don’t think the bad assets are infectious. Nor do we think this re-arrangement increases or changes values or facilitates the valuations of the assets behind the banking system.  What the change does do is make ever-so-clear what the priorities are in an insolvency.  Note that in this re-arrangement, the debt claims, including the short-term commercial paper, are direct liabilities of the bad bank.  If the bad bank cannot re-fund its commercial paper one morning, the bad bank must be re-organized.  Some of its claims must be turned into equity.  This is the standard sort of Chapter 11 re-organization.

Thus, what this re-organization of the ownership claims does is show how easy and orderly a garden-variety Chapter 11 reorganization of a large bank could be, and how unnecessary it is to throw additional public money into insolvent institutions.  And thinking about public money in insolvent institutions, consider that Citi’s present market cap of $11 billion (as of the close on Friday, 20 Feb 2009) includes the Treasury’s (that is, the public’s) investment of $50 billion in preferred stock (10/28/2008 and 12/31/08), and asset guarantees (1/16/09) are unambiguously underwater.

At the most practical level, the advantage of the good-bank/bad-bank separation is to prevent the emergency that would occur if a large bank threatened insolvency. For Citibank, which has large amounts of uninsured deposits, which would remain in the good bank, the first symptom would be a run on the bank by those depositors. At that time, the pressure for offsetting the emergency by an injection of public money would be overwhelming. By contrast, no run would occur on the heavily capitalized good bank in our example. Reorganization could proceed peacefully while the good bank went about its banking business. The claims of the shareholders and bondholders, which are inferior to those of the depositors, can be sorted out without interfering with the operation of the bank.

Much thinking about bank policy takes an old-fashioned point of view by assuming that a bank finances all of its assets through deposits. The good-bank/bad-bank separation has no advantage in that traditional setting. But for a bank that is mostly financed by non-deposit borrowing, moving the non-deposit liabilities to the bad bank has an advantage in dealing with insolvency.


Employment crosses the line

February 7, 2009

With the January figure, payroll employment has crossed the line of the worst previous recession, the one beginning in July 1981, rescaled to the size of today’s labor force. One more bad month and the current recession will become the worst on record, apart from the Great Depression.


The Fed contracts

February 4, 2009

The Fed has indicated that it plans to pursue a policy of quantitative easing, that is, expanding its portfolio by borrowing in financial markets at low rates and investing the proceeds in higher-yielding private investments. The Fed’s acquisition of large amounts of Fannie and Freddie’s debt was a notable success of that strategy, as it lowered their borrowing rates and thus lowered mortgage rates. But, as the graph below shows, the Fed has engaged in quantitative tightening over the past month, reducing its borrowing and reducing its holdings of higher-yielding investments. The line labed “Other assets” measures that type of investment. So far, no explanation for the Fed’s announcements of moving in an expansionary direction while actually contracting.


Venture Capital in the Recession

February 2, 2009


While plant and equipment spending for the entire economy was down in the fourth quarter, it only fell by 4.4 percent from its level in the fourth quarter of 2007.  By contrast, venture capital investment was down by 31 percent.  In the fourth quarter of 2007, $8.1 billion was invested in 850 venture deals.  In the fourth quarter of 2008, $5.6 billion was invested in 790 deals. The number of deals is not abnormally small—over the last five years, the fourth quarter has seen 750 deals on average, but the total invested is down a lot from last year.


This bodes ill for the 2000 vintage of venture capital.  Compare and contrast two vintages of venture:  1992 and 2000.  First note that we look at venture vintage years a bit differently from how other analysts look at them.  Others look at the year a venture fund originated.  Our vintages represent a set of companies that got their first round of venture funding in a given year.  In our analysis, the 1992 vintage year is all companies who received a first round of venture funding in 1992. 


1992 was a good year and a typical year for venture capital. By our reckoning, 1995 was the best year.  Reason is that the companies were mature enough to exit (go public or be acquired) at a time when the stock market was high.  Here’s the evolution of that 1992 vintage: 



To produce this chart, we add up all the investments made in the 1992 vintage companies, both in 1992 and all subsequent years.  We benchmark that total at 100 to make years easy to compare.  The blue line shows the total capital to be invested in the companies, and it declines over time as investments are made.  The red line shows the value of portfolio companies that are in venture portfolios, only from the 1992 vintage.  The black line shows cumulative distributions to investors from acquisitions and IPOs.  (This chart does not follow companies after they go public.  We figure there are already enough folks tracking public companies.)   Then there is the purple “total” line, the sum of capital that will be invested, value carried in private companies, and distributions to investors. Notice how the total line slopes upward most steeply in the early years, typically years 3-6, as the best investments exit.  The best exits are among the earliest, it seems for no more profound reason than the best ideas are evidently the best fairly early.  As time passes, the additional exits are middling.  Chances are, after 7 or 8 years, that if a company has not done something great for its investors as yet, it is not going to. This is a bit less true for the life science sector of venture than for information technology. But then, life science is the less risky sector of venture capital by every metric.

 Note how the rising tide of the stock market raised all values in 2000. For venture capital 2000 was an extraordinary year, as $95 billion was invested in venture companies during that year (vs. $30 billion in 2008).  But the prices paid for investments in that year were high.  Here’s the vintage year evolution for 2000: 



For 2000, distributions (from IPOs and acquisitions) are still below the amount invested. The “total” value line (purple) is above 100 near its end only because of the value at which portfolio companies are being carried.  With the recent blow to asset prices, these must fall.  Thus, it looks like 2000 will be the first year that venture capital as a whole will fail to return capital to investors. 


Note that these figures are before  the general partners’ take (a percent of invested funds and a fraction of capital gains), and are thus gross results, not net results to investors.  They are compiled from the database at Sand Hill Econometrics,built from data from Thomson, DowJones VentureOne, and a lot of our own research to fill in holes for valuations, shutdowns, and acquisition values.  We believe that this data comprise very close to the universe of venture-funded companies from 1989 to the present.

Fiscal stimulus: More needed?

February 1, 2009

The House has passed HR1 and the Senate is working on a counterpart fiscal stimulus bill. We will discuss HR1 for the sake of specificity and in the belief that the final law will be fairly similar to HR1.

The table below shows the Congressional Budget Office’s estimates of the spending increases and tax cuts under HR1. The government’s fiscal year ends on September 30, so the first period is the 8 months from now until the end of September. One of the major concerns about the program is the lag in ramping up the stimulus. Recall that the economy will be about $900 billion below trend in the middle of the year, but spending programs up to then will only pay out $107 billion and taxes will have been cut by $63 billion. The stimulus is much larger in the following 12 months, starting in the fall of this year, when spending will be up by $236 billion and taxes cut by $120 billion. Then in the following fiscal year, 2010 to 2011, the stimulus falls to roughly half its peak level. Effects in later years are tiny.


 The spending increases under HR1 are wonderfully diversified. We had been concerned that the program would involve large amounts of spending in certain favored, narrow areas, and that the supply of the corresponding products would have bottlenecks. In that case, the direct employment effects would be smaller and the spending would have driven up the incomes of some lucky workers. But in 2009 and 2010, broadband gets only $250 million, school building repairs $9 billion, and health information technology $500 million, to mention three areas that the new administration had been singling out earlier as especially suitable. Though the spending side of the program has been widely criticized as diffuse, we see that characteristic as a virtue.

The biggest piece of the tax side of the program is a conventional tax rebate, similar to the one in the summer of 2008. Because it would be refundable-that is, it could result in a taxpayer receiving money back rather than paying less tax-the government accounts for it partly as spending (the refunds) and partly as reduced taxes. The total amount is $144 billion for 2009 and 2010. On the business side, speeding up depreciation and allowing earlier recovery of tax benefits from operating losses would result in a drop in tax collections of $90 billion, to be made up with higher collections in subsequent years-this is an interest-free loan program more than a tax cut.

Effects of fiscal stimulus

How much does real GDP respond to the two forms of fiscal stimulus, government purchases and tax cuts? The incoming Obama administration released the Romer-Bernstein study of fiscal stimulus to answer the question. The table below shows the main conclusions. When the government increases its purchases of output, say concrete for new bridge, GDP rises immediately by about the amount of the purchase (1.05 in the first quarter) and then rises to about 1.6 times the amount of the purchases. The calculation assumes that the flow of purchases remains higher for the whole 16-quarter period shown in the table. For tax cuts or other programs that give families more purchasing power, the immediate response is more like half the amount the government spends; the other half goes into savings. In both cases, the multipliers rise as time passes because of follow-on effects-increased spending by those influenced by the immediate effects.


 Fiscal Multiplier Estimates from the Romer-Bernstein study of fiscal stimulus

The measurement of these multiplier effects is intensely controversial. Having followed this area of macroeconomics carefully for many years, we believe that both columns overstate the responses somewhat, especially in the case of tax cuts. All macro models agree that increases in government purchases increase real GDP. That multiplier is probably around one, we believe, but we would not rule out 1.5. Most macro models place the tax-cut multiplier quite a bit lower, and we agree. Here we are referring to tax cuts that hand families more purchasing power, such as the one n the summer of 2008. Later we discuss tax changes that give focused incentives for immediate spending, where the effect could be much larger.

Last summer, consumers enjoyed a moderate tax cut paid out to most people as an immediate rebate. The experience demonstrated that the federal government is capable of a speedy fiscal action, but it also showed that consumers don’t go out and spend rebate money when they receive it. Instead, just as the received theory of consumption predicts, consumers smooth the spending increase, reserving most of it for future consumption, by saving it or using it to pay down debt. It raises their standard of living more than their immediate spending. It’s good news that most Americans do not have their backs to the wall financially, but it makes an income- tax rebate ineffective as a stimulus concentrated at the time most needed. Rather, rebates result in consumption increases spread well into the future. We discussed John Taylor’s graphic analysis of the tdax cut in an earlier post.

Effect of Government Purchases

HR1’s focus on government purchases raises the natural question of the effect of purchases on total GDP. Does government spending stimulate other categories of spending, especially consumer spending? Or does government spending displace other categories, so GDP rises by less than the amount the government spends? What does research say about the left column in the table above? An earlier post addressed this question.

What could have been done

The most desirable time-concentrated stimulus would raise consumer spending during 2009 without changing it much in future years. The government lacks the power to compel consumers to spend more, so it must rely on a incentives. Britain has just cut its sales tax (VAT) by 2.5 percentage points for the next 13 months to provide a small incentive concentrated in the period when the recession is expected to be most severe.

An important article in the American Economic Review in September 1986 by James Poterba, Julio Rotemberg, and Lawrence Summers demonstrated the high potency of a sales-tax cut as a stimulus. They studied the reverse situation, where a big increase in the British sales tax caused a severe recession.

Because the U.S. lacks a national sales tax or VAT, the logistics of a temporary consumption subsidy would be a little more complicated. The Kotlikoff-Leamer proposal would operate through state sales taxes. All but a few small states have sales taxes-the exceptions are Alaska, Delaware, Montana, New Hampshire and Oregon. Under the plan, the federal government would buy out sales taxes for the period of the needed stimulus, say the year 2009. The states without sales taxes or with low sales taxes would receive comparable federal funds to cut other types of taxes. Sales tax revenue is currently $440 billion per year, so the proposal would cost around half of what the Obama administration appears to be planning to spend on stimulus.

The plan needs to take effect soon after it is announced. The announcement will cause consumers to defer purchases until the tax cut takes effect. Similarly, toward the end, they will accelerate purchases and then buy less after the sales tax resumes. Phasing in the resumption might be a good idea.

Critics of consumer subsidies point out that some consumer goods are imported, so the stimulus benefits the producing country, not the U.S. Often that country is said to be China, though Japan is also a major source because of its dominance of the world car market. Given that the U.S. has pledged to engage in joint stimulus with its major trading partners, including China, the stimulus to the import source countries is entirely appropriate. But imports of consumer goods are only 18 percent of consumer spending on goods (excluding services), so the leakage into import stimulus is not a major consideration anyway.

We feel that a temporary elimination or reduction in sales taxes would be an effective stimulus to consumer spending, concentrated in the period when it is needed most and phased out later. It should be part of the stimulus plan.

Tax cuts to reduce the cost of labor or improve incentives to work

If the objective of a stimulus is to lift employment, why not operate directly on that margin, by improving incentives? The Senate is considering a program called Making Work Pay (see the Brookings Institution evaluation) as part of its version of the stimulus bill. This program would buy out the first $500 of the payroll tax that all workers have to pay, even those who do not pay income tax. This program has a definite effect on the incentive to work for those without jobs and for those who are currently paying less than $500 in payroll tax (those earning less than about $8,000 per year), but for most workers, the $500 would be just like a standard tax cut, without any special incentive effect.

In a standard supply-and-demand diagram, with the pre-tax wage on the vertical axis, giving an improved incentive to workers shifts the supply curve to the right-people are willing to work more hours at a given pre-tax wage because the tax that they used to pay has been removed, the equivalent of an increase in the wage.  The market clears at a point down the demand curve, with higher employment. That is the hope of a plan to cut the wage tax paid by workers. Notice that the pre-tax wage as to fall to stimulate demand. If the wage is sticky, nothing happens-workers would like work harder, but that message does not get passed on to employers through a lower pre-tax wage. Thus another reason that Making Work Pay may not stimulate much job-creation is that the needed decline in pre-tax wages won’t occur.   

Why not overcome this problem by giving the incentive to the employer? The federal government has a completely straightforward way to do this, because it levies a 7.65 percent rate on employer payrolls for virtually all employment. The tax yields just under $500 billion per year. Elimination of the tax for the year 2009 would provide a substantial incentive to employment concentrated during the year. As with the temporary removal of sales taxes, both anticipation effects and ending effects would occur. See the Bils-Klenow proposal.

One problem with the employment stimulus is that the funds go in the first instance to the owners of businesses and not to consumers generally. Ownership is highly concentrated in the U.S., so the distribution of the immediate benefits is skewed. By contrast, a worker or consumer rebate can be directed to lower-income consumers because it is part of the income tax.

If the response of employment to the payroll tax cut were strong enough, its contribution to the incomes of workers might be enough to overcome the disadvantage of its business-subsidy character. If business hiring responded aggressively to the subsidy, profits would decline as business put more output on the market and the winners would be workers rather than owners. Everything turns on the strength of the employment effect. But, alas, the strength of the effect depends on one of the most unsettled issues in macroeconomics, the role of supply increases in raising the quantity of output produced in the short run. One line of thought treats this issue just as one would in a standard market, where an increase in supply raises the quantity sold by the principles of the standard supply-and-demand diagram of elementary economics. Another line believes that special principles operate in the short run that makes demand the controlling factor-an increase in supply has little effect on the quantity sold in the short run in this view. We are among the few economists who regard this issue as still open. Most are doctrinaire believers in one or the other view. We are sufficiently concerned about the potential validity of the demand-limiting view that we are reluctant to state with confidence that a supply-based stimulus such as the payroll-tax suspension would have a large effect on employment.

Will the economy need more stimulus after the current program goes into effect?

The answer seems obviously yes. Current forecasts were made with pretty good knowledge of the nature and magnitude of the Obama stimulus program and they say that the economy will be in deep recession through the middle of the year and below trend with excess unemployment for several more years. Further, from Tables 2 and 3, it is clear that the increases in purchases and increases in consumer purchasing power from the program, after applying the Romer-Bernstein multipliers, would have effects quite small in relation to the GDP shortfall of close to a trillion dollars a year.

Conclusions on stimulus

We foresee a mixture of stimulus policies for the coming year. Monetary policy can contribute by driving down mortgage and other interest rates. Income-tax rebates seem to have little support and would probably have relatively small effects within the year, with undesirable continuing effects in later years. We are enthusiastic about removing sales taxes and then brining them back gradually, to stimulate immediate consumptions spending. We are not sure that an employment stimulus from a reduced business payroll tax would raise employment enough to be a contender as a stimulus. Thus the sales-tax buyout seems to be the best way to spend the bulk of future stimulus dollars.