PrefBlog

OSFI has republished a puff-piece written for Central Banking magazine, titled Lessons for Banking Reform: A Canadian Perspective, by Carol Ann Northcott & Graydon Paulin of the Bank of Canada and Mark White of … OSFI.

Credit for Canada’s performance throughout the crisis is given to:

• High levels of capital
• A rational mortgage market
  • relatively low Loan-to-Value
  • Recourse to borrower
  • Non-deductability of interest
• Assets-to-Capital (ACM) multiple control
• lack of competition from shadow banks
• The wise and beneficient supervision of those sadly underpaid geniuses (genii?) at OSFI

Not much meat on these bones, frankly. I would have been much more interested in a solid analysis of just WHY we were so lucky. Why weren’t the banks up to their necks in sub-prime paper, like everybody else? Was it the ACM? Was it because Canadian banking is such a profitable rent-extraction machine that banks didn’t need to lever up on Sub-prime at LIBOR+50? I find the idea that “Canadian Bankers are Smart” rather difficult to swallow. We nearly went bust in the MBA crisis of the 1980’s … we’ll find something else soon, don’t fret.

And why are we so highly capitalized, anyway? It has been very useful in the downturn, there’s no denying that … but what are the net, through-the-cycle cost/benefits of tying up a lot of capital in the banking system?

Gareth Campbell writes an interesting essay on VoxEU, The railway mania: Not so great expectations?, in which he uses the British Railway Mania of the 1840’s to consider the problem of deciding ex ante whether there is a bubble in asset prices:

Can financial crises be averted by identifying and dealing with overpriced assets before they cause instability? This column argues that during the British Railway Mania of the 1840s, railway shares were not obviously overpriced, even at the market peak, but prices still fell dramatically. This suggests that extreme asset price reversals can be difficult to forecast and prevent ex ante, and the financial system always needs to be prepared for substantial price declines.

Assiduous Readers will be well aware of my view that market timing is difficult to do … really, really, really difficult to do … impossible. Comparing one share in a bank to another share in another bank is relatively easy. Comparing one share in a bank to cash … can’t be done.

This concept has become important in that there is a move afoot to give central bankers a mandate to time the markets:

The instability that has followed the bursting of the housing bubble has led to a renewed discussion about what can be done to prevent the recurrence of financial crises. Cecchetti et. al (2000) have suggested that monetary policy should be tightened when regulators believe assets are overpriced, in an attempt to deflate a suspected bubble before it bursts. However, Bernanke (2002) and Mishkin (2008) have argued that this proposal is not feasible, partly because mispricing is difficult to identify ex ante. Several pieces of academic research have provided a justification for this position by suggesting that assets were not obviously mispriced prior to market crashes in certain historical episodes, such as the Tulip Mania of 1636 (Garber, 2001), the German stock market boom of 1927 (Voth, 2003), and before the Wall Street Crash of 1929 (Donaldson and Kamstra, 1996).

It’s an interesting piece and ties in with my fears regarding the too-popular dividend growth magic formula, that I fear will increasingly lead to unwise decisions being made regarding dividend payouts and lead to very crowded trades when the hard decisions finally become effective.

First he shows the market indices for All Railways, Established Railways and Non-Railways:

Click for big

Then he shows how dividends rose and fell through the period:

Click for big

Then he compares the dividend yield on railways with non-railways

This involved performing a regression for each week of the sample using the dividend yield of a company as the dependent variable and using a dummy variable, which equalled 1 if a company was a railway, as the independent variable. The coefficient of the railway dummy in each week is plotted with ±1.96 standard errors in Figure 3.

Click for big

and finally:

I have also extended the analysis of the overpricing or underpricing of railway shares by including as independent variables the dividend growth that the company went on to experience during the next three years. After controlling for this growth, the apparent overpricing of the railways during the boom in prices is almost entirely eliminated. The railways appear to have had a significantly lower dividend yield, after accounting for short-term dividend growth, on just two weeks during the entire period, as shown in Figure 4.

Click for big

He concludes:

Regulators may be able to effectively intervene if they have greater foresight than other market participants but, as Bernanke (2002) has argued, this is a very questionable assumption. Without perfect foresight by regulators, which would allow restrictions to be imposed only when necessary, there would have to be tighter regulation in all periods, and the costs of such restrictions would need to be carefully weighed against the potential benefits.

It may be better to focus efforts on ensuring stability when the next asset price bust does occur. More attention should be directed to the consequences of sustained declines in asset prices, as capital requirements which are based on short-term market risk may provide inadequate protection against persistent and longer-term falls. It may be useful to introduce some long-term stress testing, which would examine the consequences if the largest peak-to-trough asset price movements in history were to be repeated. If the financial system could not endure some of these historical experiences, then it may need to embrace tighter regulation or restructuring.

I continue to like the idea of dynamic provisioning, in which a bank’s “recent” assets would carry a higher risk weight than “old” assets, penalizing recent growth to a predictable degree (in addition to the unrelated idea of penalizing excessive size, inter alia). Dynamic provisioning has, to some extent of equivalency, been implemented by the FDIC.

I like those ideas a LOT more than telling Bernanke that the non-bonusable nature of his position makes him an infallible market timer.

Mikhail Chernov & Philippe Mueller: The Term Structure of Inflation Expectations:

The ten-year inflation premium declines from six to zero per cent during the post-monetary-experiment period. This decline suggests that long-run inflation expectations became more stable over time. Further, we reestimate our model every quarter and find that the long-run expectations have declined over time from 6% to 2%. The inflation persistence declined and the term structure of inflation expectations became flat over time. This evidence suggests that monetary policy became better anchored.

One implication of anchored inflation expectations is that it should be easier to forecast inflation and yields. Consistent with this prediction, we find that the model that incorporates both yields nd surveys dominates in out-of-sample forecasting of both inflation and yields. These results lead us to conclude that information in surveys is extremely important for establishing the links between inflation expectations and yields.

Figure 4 shows the time-series of the inflation expectations at multiple horizons. These expectations are computed from AO. In contrast to the survey forecasts in Figure 1, these objective, or marginal, expectations can be computed each period at any horizon.

The term structure effects are pronounced. The inflation curve becomes inverted in 1973, right before the recession, and continues to be inverted until early 1982. This period coincides with the unstable period of monetary policy during the Burns and Miller chairmanship of the US Federal Bank and the monetary policy experiment under Volcker’s chairmanship. The curve became inverted again briefly in the early part of Greenspan’s tenure from 1987 to 1991. afterwards, it had a normal, nearly flat, shape.

Click for big

The out-of-sample analysis of the model suggests that monetary policy became more effective over time. The long-run expectations are anchored at about 2%. The term structure of inflation expectations has flattened out over time. This suggests that the arrival of new data does not affect long-run expectations much, perhaps because the monetary policy is expected to address all short term fluctuations successfully.

Christopher Reid, Frédéric Dion and Ian Christensen, Real Return Bonds: Monetary Policy Credibility and Short-Term Inflation Forecasting [broken link fixed 2023-1-21], Bank of Canada Review, Autumn 2004:

Chart 5 shows two proxies of long-run inflation uncertainty. The first is a measure of the disagreement among forecasters who responded to the Watson Wyatt survey, calculated as the difference between the upper and lower quartiles of reported inflation expectations at the 4- to 14-year horizon. The second measure is inflation uncertainty over a 5-year forecast horizon derived from a GARCH model developed by Crawford and Kasumovich (1996).

Côté et al. (1996) suggest that the increase in the BEIR [Break-Even Inflation Rate, Nominals less RRBs] in 1994, which was not accompanied by a similar move in survey measures, may reflect an increase in the inflation-risk premium. If changes in the premium for inflation uncertainty are an important factor in explaining movements in the BEIR, then sharp movements in these proxies should be associated with similar movements in the BEIR. Yet both measures fail to indicate a rise in inflation uncertainty in 1994 or a significant decline in 1997. Crawford and Kasumovich’s measure of inflation uncertainty fell dramatically during the 1980s but has been relatively stable since 1992. Similarly, survey disagreement fell between 1991 and 1994 but was relatively stable afterwards. The simplest explanation is that deviations of the BEIR from survey measures of inflation expectations are the result of some phenomenon other than changes in uncertainty regarding inflation.

Grishchenko, Olesya V. and Huang, Jing-Zhi, Inflation Risk Premium: Evidence from the TIPS Market (December 11, 2008).

“Inflation-indexed securities would appear to be the most direct source of information about in°ation expectations and real interest rates” (Bernanke, 2004). In this paper we study the term structure of real interest rates, expected in°ation and inflation risk premia using data on prices of Treasury Inflation Protected Securities (TIPS) over the period 2000-2007. The estimates of the 10-year inflation risk premium are between 11 and 22 basis points for 2000-2007 depending on the proxy used for the expected inflation. Furthermore, we find that the inflation risk premium is time varying and, specifically, negative in the first half (which might be due to either concerns of deflation or low liquidity of the TIPS market), but positive in the second half of the sample.
…
This paper represents perhaps the first attempt to estimate the inflation risk premium directly using the prices of Treasury Inflation Protected Securities (TIPS). Using the market data on prices of TIPS over the period 2000-2007, we find that the 10-year average inflation risk premium ranges from 11 to 22 basis points. We also find that it is time-varying. More specifically, it is negative in 2000-2003 but positive in 2004-2007. The negative inflation risk premium during 2000-2003 is due to either concerns of deflation or liquidity problems in the TIPS market. There seems to be more evidence that supports the former explanation. The estimated average 10-year in°ation risk premium over the second half varies between 29 and 48 basis points, depending on the proxy used for the expected inflation. The estimates based on Blue Chips inflation forecast are the lowest (29 basis points), and the estimates based on one-year SPF are the highest (48 basis points). We also find that the inflation risk premium is considerably less volatile during 2004-2007, a finding consistent with the observations that in°ation expectations became more stable during this period, investors became more familiar with the TIPS market, and the market liquidity has gradually improved.

Our empirical results on in°ation risk premium estimated directly from TIPS should be valuable for practitioners, monetary authorities and policymakers alike because they help to assess the inflation expectations and the inflation risk premium of bond market investors.

The Inflation Risk Premium in the Term Structure of Interest Rates, Peter Hördahl, BIS Quarterly Review, September 2008:

A dynamic term structure model based on an explicit structural macroeconomic framework is used to estimate inflation risk premia in the United States and the euro area. On average over the past decade, inflation risk premia have been relatively small but positive. They have exhibited an increasing pattern with respect to maturity for the euro area and a flatter one for the United States. Furthermore, the estimates imply that risk premia vary over time, mainly in response to fluctuations in economic growth and inflation.
…
This article estimates inflation risk premia using a dynamic term structure model based on an explicit structural macroeconomic model. The identification and quantification of such premia are important because they introduce a wedge between break-even inflation rates and investors’ expectations of future inflation. In addition, inflation risk premia per se may provide useful information to policymakers with respect to market participants’ aversion to inflation risks as well as to their perceptions about such risks.

The results show that inflation risk premia in the United States and in the euro area are on average positive, but relatively small. Moreover, the estimated premia vary over time, mainly in response to changes in economic activity, as measured by the output gap, and inflation. The estimates suggest that fluctuations in output drive much of the cyclical variation in inflation premia, while high-frequency premia fluctuations are mostly due to changes in the level of inflation.

Click for big

I missed this when it came out.

The FDIC has released a rule increasing the complexity of deposit insurance premium calculations:

The final rule adds a new financial measure to the financial ratios method. This new financial measure, the adjusted brokered deposit ratio, will measure the extent to which brokered deposits are funding rapid asset growth. The adjusted brokered deposit ratio will affect only those established Risk Category I institutions whose total gross
assets are more than 40 percent greater than they were four years previously, after adjusting for mergers and acquisitions, rather than 20 percent greater as proposed in the NPR, and
whose brokered deposits (less reciprocal deposits) make up more than 10 percent of domestic deposits. Generally speaking, the greater an institution’s asset growth and the greater its percentage of brokered deposits, the greater will be the increase in its initial base assessment rate. Small changes in asset growth rate or brokered deposits as a percentage of domestic deposits will lead to small changes in assessment rates.

The Canadian approach is not nearly so nuanced since our bankers are ever so smart. In fact, they’re all equally smart, with the vast majority of assets in the system paying into the CDIC fund at the same rate, which is considered desirable. Not so in the States:

A commenting bank argued that:

Arbitrarily establishing targets for percentages of institutions that fall into a given assessment rate is inconsistent with not only the governing statute but the whole concept of risk-based pricing….

The FDIC disagrees with the commenting bank. The purpose of the new large bank method is to create an assessment system for large Risk Category I institutions that will respond more timely to changing risk profiles, will improve the accuracy of initial assessment rates, relative risk rankings, and will create a greater parity between small and large Risk Category I institutions.

Imagine that! Rewards for being better than the competition, even if only by a little bit! It’s a good thing we don’t have that sort of nonsense in Canada – it can lead to bonuses.

Since the FDIC is not Canadian, they address criticism, allowing investors and observers to take an informed view of the desirability of changes:

The FDIC received many comments arguing that brokered deposits should not increase assessment rates for Risk Category I institutions and that the brokered deposit provisions in the NPR do not account for the use to which institutions put these deposits. The FDIC is not persuaded by the arguments. Recent data show that institutions with a combination of brokered deposit reliance and robust asset growth tend to have a greater concentration in higher risk assets. In addition, there is a statistically significant correlation between the adjusted brokered deposit ratio, on the one hand, and the probability that an institution will be downgraded to a CAMELS rating of 3, 4, or 5 within a year, on the other, independent of the other measures of asset quality contained in the financial ratios method.

The Bank of Canada has released the Spring 2009 Review with the following feature articles:

• Next Steps for Canadian Monetary Policy
• Price-Level Targeting and Stabilization Policy: A Review
• Price-Level Uncertainty, Price-Level Targeting, and Nominal Debt Contracts
• Unexpected Inflation and Redistribution of Wealth in Canada

The concept of Price Level Targetting is explained in the first article:

Despite its recent successes in terms of macrostabilization, several authors have highlighted some shortcomings in the infl ation-targeting (IT) framework. Most notably, uncertainty on the price level grows with the planning horizon, since central banks with infl ation targets accommodate shocks to the price level, taking the post-shock level as given and aiming to stabilize infl ation from this level. In fact, the price level is unbounded at very distant horizons. Price-level targeting (PT) mitigates this uncertainty by committing central banks to restore the price level to a preannounced target following shocks. PT is frequently described as a departure from IT’s prescription for letting “bygones be bygones.”

Frankly, I didn’t find this issue particularly satisfying; there are necessarily many assumptions embedded in the papers. There is the prospect of lowering the term risk premium (flattening the yield curve) with Price Level Targetting, but on the other hand it’s asking rather a lot from the Central Bank, which will have to overcompensate for transient shocks rather than concentrating on getting things back to normal.

This seems to be a hot topic, so I’ll post a reference to Transparency and the Corporate Bond Market by Hendrik Bessembinder and William Maxwell of the universities of Utah and Arizona, respectively n.b.: link updated 2011-4-30. Old link no longer works:

The introduction of TRACE to the bond
market provides a rare opportunity to assess the effects of a substantial increase in transparency.
…
While over-the-counter corporate bond trades tend to be large, they also tend to be infrequent. Edwards, Harris, and Piwowar (2007) report that individual bond issues did not trade on 48 percent of days in their 2003 sample, and that the average number of daily trades in an issue, conditional on trading, is just 2.4 Corporate bonds trade infrequently even compared to other bonds. Although Table 1 shows that they comprise about 20 percent of outstanding U.S. bonds, corporate bonds account for only about 2.5 to 3.0 percent of trading activity in U.S. bonds in recent years, as shown in Table 3.
…
That execution costs for bonds decline with trade size may reflect in part that asymmetric information regarding issuing firm fundamentals is relatively unimportant for bond valuation. It could also reflect the absence of an inexpensive centralized system for processing small bond transactions. Or the higher execution costs for small bond trades could reflect the extraction of rents by better-informed bond dealers from relatively uninformed retail bond traders.
…
Well-functioning security markets provide investors with liquidity. However, the term “liquidity” is a broad and somewhat elusive concept, used to describe multiple properties of trading in security markets. For example, Kyle (1985) notes that liquidity can include “tightness,” which is the cost of completing a buy and sell transaction in a short period of time, “depth,” which the size of the buy or sell order required to move market prices by a given amount, and “resiliency,” which is the speed with which prices recover from a random shock in buy or sell orders. Alternately, practitioners sometimes use the word liquidity to describe the ease of transacting.
…
Empirical evidence on the introduction of transaction reporting in corporate bonds has been the subject of countless articles in the trade press and at least three articles published in refereed academic journals: Bessembinder, Maxwell and Venkataraman (2006), Edwards, Harris, and Piwowar (2007), and Goldstein, Hotchkiss, and Sirri (2007). Although the three studies use notably different samples and research designs, all three conclude that the increased transparency associated with TRACE transaction reporting is associated with a substantial decline in investors’ trading costs.
…
Bessembinder, Maxwell and Venkataraman (2006) also examine how transparency affects the competitive environment of the dealer market. They hypothesize that in an opaque market the largest dealers enjoy an informational advantage, but that this informational advantage is mitigated in a transparent market. Consistent with this reasoning, they report that in their sample the concentration ratio of trades completed by the largest 12 dealers falls from 56 percent pre-TRACE to 44 percent post-TRACE.
…
Market participants with whom we spoke, including both dealers and the traders at investment firms who are their customers, were nearly unanimous in the view that trading is more difficult after the introduction of TRACE. Whereas it may have previously been possible to complete a sizeable bond purchase with a single phone call to a dealer who held sufficient quantities of the bond in inventory, the post-TRACE environment may involve communications with multiple dealers, and delays as the dealers search for counterparties. A bond trader with a major insurance company told us that there is less liquidity, in that market makers carried less “product,” and it has become more difficult to locate bonds for purchase in the post-TRACE environment. A bond trader for a major investment company responded to the publication of Bessembinder, Maxwell, and Venkataraman (2006) by sending the authors an unsolicited e-mail stating: “I want to be able to execute a trade even if a bond dealer does not have a simultaneous counterparty lined up…. [T]oo much price transparency reduces dealers’ willingness to commit capital…. [T]he focus on the bid-ask spread is too narrow, and a case of being penny-wise and pound-foolish.”
…
One way to circumvent TRACE, which applies to publicly-issued bonds, is for a firm to issue privately placed bonds (sometimes referred to as Rule 144a securities, for the section of the Securities Act of 1933 that provides exemption from registration requirements). … In 2001, before TRACE, “144a for life” bonds were 7.3 percent of dollar volume and 9.6 percent of issues. The percentage of dollar volume in “144a for life” bonds jumped to 27.8 percent in 2003, the first full year after TRACE initiation, and grew to 39.8 percent in 2004, before declining to 16.9 percent in 2006.
…
Also consistent with a shift towards alternative asset classes, the credit default swap market experienced phenomenal growth in recent years relative to bonds. Table 6 reports on outstanding notional principal in these credit default swaps, which grew from $919 billion in 2001 to $34.4 trillion in 2006. One dealer suggested to us that, prior to TRACE introduction, ten times as much capital was allocated to corporate bond trading than to credit default swaps, but that the ratio has now been reversed.
…
To the extent that the shift to
privately placed bonds and bank loans was initiated by corporate borrowers, and in response to
TRACE, it suggests that the net costs of TRACE may exceed the benefits….Alternately, the shift to private markets could simply reflect agency issues if issuers failed to fully anticipate the potential effect of illiquidity on issue prices and underwriters and lenders persuaded corporations to issue private securities that could be traded more profitably.
…
A number of industry participants told us that bond dealers have either reduced expenditures for research regarding bond valuation, or have stopped providing the research to customers, instead using it for proprietary trading. A trader for a major market-making firm noted that the easiest way to cut expenses in the wake of lower bid-ask spreads was to reduce the number of analysts on the payroll. Some bond dealers, including Citibank, no longer provide external research on the corporate bond market.
…
The primary complaint against TRACE, which is heard both from dealer firms and from their customers (the bond traders at investment houses and insurance companies), is that trading is more difficult as dealers are reluctant to carry inventory and no longer share the results of their research. In essence, the cost of trading corporate bonds decreased, but so did the quality and quantity of the services formerly provided by bond dealers.

A recent draft paper by Alan V.S. Douglas, Alan G. Huang & Kenneth R. Vetzal (all of the School of Accounting & Finance, University of Waterloo), Cash Flow Volatility and Corporate Bond Yield Spreads demonstrates that there is pricing information in firms’ cash flow volatility that is not captured by more usual metrics:

Control variables were

• Issuer Credit Rating
• Years to Maturity
• Coupon Rate
• Liquidity
• Debt Servicing Ability
• Leverage
• Equity return volatility
• Term Structure Level
• Term Structure Slope

A fundamental determinant of firm value is cash flow. Accordingly, the uncertainty or volatility associated with cash flow should be reflected in default probabilities and bond yield spreads. This paper tests the cross-sectional, inter-temporal and overall relationships between volatility and spread using both expected and historical measures of cash flow volatility. We find that cash flow volatility is economically significant in explaining yield spreads. Expected cash flow volatility explains 51 basis points of yield spread in the univariate regression, and 17 basis points after controlling for the commonly used spreadinformative variables. Historical cash flow volatility explains yield spread with a similar magnitude. Importantly, we show that the cash flow volatility effect is robust to the closest proxies of asset volatility used in the literature, namely, stock return volatility, accounting earnings volatility, and analyst forecast dispersion of earnings. Our study highlights the importance of cash flow uncertainty risk in pricing corporate bonds.

This paper is an interesting extension of the Merton Model; it would be most interesting to see how this measure of risk has evolved in importance over time.