A Balance Sheet Quirk At Bank Of America

| About: Bank of (BAC)

Summary

Repo 105 was an accounting technique used by Lehman Brothers.

The technique allowed Lehman to pretend to sell off assets and pay down debt, thus making it look safer than it actually was.

In early 2010, Australian hedge fund manager John Hempton indicated that Bank of America engaged in similar practices before the financial crisis.

I later discovered in late 2011 that Bank of America may have still been engaging in these practices in 2010 and 2011.

I reexamined Bank of America's balance sheet recently, and it still shows some evidence of unusual asset trends.

Introduction: Repo 105 and Lehman Brothers

Repo 105 was an accounting technique used by Lehman Brothers to hide how much money it borrowed. The company would use assets as collateral for short term loans. It then used the loans to pay off other debts. The company got $100 in loans per $105 in collateral, hence the name "Repo 105."

What was unique about Repo 105 was that it let Lehman look safer than it actually was. For the duration of the loans, the company could say it had sold the assets used as collateral. It could do so even though it had to buy them back at the end of the loan just a few days later, generally with newly borrowed money. For those few days, though, the company could temporarily "park" the assets on another company's balance sheet-for a fee, of course.

Lehman used Repo 105 to look make its balance sheet look more attractive in its filings to the Securities and Exchange Commission, or SEC. Right before making its filings, the company would use Repo 105 to pretend that it had sold assets and reduced debt. This reduced Lehman's leverage ratio (the ratio of its assets to its equity), as calculated from the company's SEC filings. Because the leverage ratio is a key measure of a bank's safety, this made the company look safer than it actually was.

Unsurprisingly, Lehman became more aggressive in using Repo 105 as it ran into trouble in 2007 and 2008. According to a 2010 Financial Times article by Tracy Alloway, the bank "even [breached] its own internal cap on the Repo's use (about $22bn as of summer 2006)." The definitive report on Lehman's failure was published in 2010 by the company's bankruptcy examiner, Anton Valukas. According to the Valukas report, "Lehman temporarily reduced its net balance sheet at quarter‐end through its Repo 105 practice by approximately $38.6 billion in fourth quarter 2007, $49.1 billion in first quarter 2008, and $50.38 billion in second quarter 2008."

Lehman Brothers' use of Repo 105 was important enough that Valukas devoted an entire volume of his nine volume report to it. As the Valukas report puts it, Repo 105 "[created] a materially misleading picture of the firm's financial condition in late 2007 and 2008."

John Hempton Reveals Repo 105 Type Transactions At Bank of America

When the Valukas report came out in 2010, John Hempton wrote about it on his blog, Bronte Capital. Hempton, an Australian hedge fund manager and one of my favorite bloggers, wrote a March 2010 blog post titled "Repo 105's antecedents: Ken Lewis." In the post, he noted that "Repo 105 type balance sheet faking was 'an old trick' and well known to anyone who cared to read balance sheets (very) carefully."

He showed this with the example of Bank of America's (NYSE:BAC) 2006 balance sheet. In his words, Bank of America's fiscal year 2006 "end period assets were always lower than the average assets." In other words, the company always had fewer assets at the end of a quarter than it did on average during the quarter. Hempton explained this was because the company parked assets off balance sheet before reporting its results at the end of each quarter.

The effect of this can be seen below:

Source: Bank of America 2006 Press Releases (here and here), via Bronte Capital

Bank of America's average total assets were higher than its end of quarter total assets for every quarter in 2006:

Source: Bank of America 2006 Press Releases, via Bronte Capital

As John Hempton put it in his blog post, such results showed how "Bank of America was parking its assets off balance sheet at the end of every quarter…and had been obscuring the fact."

Another Way of Considering Bank of America's Actions

Hempton compared end of quarter assets to average assets to argue that Bank of America used Repo 105 type transactions to reduce its end of quarter asset totals. However, in my mind, there is a potential issue with this technique. If a bank's assets are falling, then its average assets in a quarter will be higher than its end of quarter assets. This can be seen in this chart:

This will happen even if the bank isn't intentionally reducing its end of quarter assets to make its balance sheet look better.

Obviously, this wasn't an issue with Bank of America in 2006-its assets were rising over time. However, this is a potential weakness with Hempton's method of comparing average and end period assets.

However I think there is a way to compensate for this issue, so that you can detect quirks in a bank's balance sheet in both rising and falling asset environments. That is by considering the company's expected average total assets in a quarter. As I see it, a bank's assets should change roughly linearly within quarters. For example, let's say a bank goes from 1,410 to 1500 billion dollars in assets in a quarter. This is an increase of $90 billion. Because a quarter is about 90 days long, that means that the bank's assets should increase by about a billion dollars a day ($90 billion/90 days).

If a bank's assets increase linearly like this, its expected average quarterly assets should be halfway between its end of quarter total assets and its total assets at the previous quarter's end:

Obviously, this is an extremely simplified version of how a bank's assets increase or decrease in a quarter. It doesn't take into account seasonality. People take out more loans in certain months. Similarly, they are more likely to pay off debt in some seasons. For example, loan balances for lenders fall in the spring as borrowers pay off debts using tax refunds. It also doesn't take into account day to day fluctuations in assets. People take out more loans on certain days.

However, I believe such fluctuations usually cancel each other out within a quarter. Thus, for a given quarter, the bank's expected average assets should be about halfway between its end of quarter assets and its end of quarter assets for the previous quarter.

If you apply this belief to Bank of America's 2006 results, the company's expected average assets versus end of quarter assets look like this:

Source: Bank of America 2005 Annual Report, as well as 2006 Press Releases via Bronte Capital

Instead, Bank of America's 2006 asset numbers look like this:

Source: Bank of America 2005 Annual Report, as well as 2006 Press Releases via Bronte Capital

As you can see, the company's average assets were much higher in every quarter than its expected average assets, as well as its end of quarter assets.

On the one hand, it is unsurprising that Bank of America's actual average assets differed from its expected average assets. As I've noted, there are many reasons why a bank's assets fluctuate. Thus, it might seem unrealistic to model a bank's asset totals as increasing or decreasing linearly within a quarter, so that its quarterly averages fall halfway between its ends of quarter asset totals.

And yet, when you examine other large banks, their average quarterly assets do fall surprisingly close to their expected values. For example, these are Wells Fargo's (NYSE:WFC) average and end of quarter assets for the same time period:

Source: Wells Fargo Quarterly Filings

As you can see, in four of the five quarters (Q4 2005 as well as Q2 through Q4 2006), the company's average assets are close to its expected average assets.

This can also be seen in JPMorgan Chase's (NYSE:JPM) results from the same time period:

Source: JPMorgan Chase Q1 2006 and Q4 2006 Earnings Release Financial Supplements

As with Wells Fargo, JPMorgan Chase's average assets don't match its expected average assets perfectly. However, they are still generally similar.

Indeed, if you compare average assets to expected average assets for all three banks, you see that Bank of America is a clear outlier:

Source: Aforementioned Bank of America, Wells Fargo, and JPMorgan Chase Filings

Bank of America's average assets, as a percentage of expected average assets, are an anomaly because they are always much higher than those of Wells Fargo and JPMorgan Chase. However, they are also an anomaly because its average assets, as a percentage of expected average assets, are anomalously high in absolute terms. Wells Fargo and JPMorgan Chase sometimes have lower average assets relative to expected average assets. This makes sense, given the seasonal variations I described earlier. In contrast, Bank of America's average assets are always elevated relative to expected average assets. Thus, assuming Wells Fargo and JPMorgan Chase reflect how a bank's assets should look, this implies something unique about Bank of America.

In conclusion, I think finding the expected average assets and comparing it to the actual average assets has two advantages. First, it compensates for the differences between average and end of quarter assets which occur when a bank's assets are rising or falling. Second, it gives another way of showing that a bank's average quarterly assets are unusual. John Hempton's analysis persuasively shows that Bank of America's average assets were unusually elevated in the quarters of 2006 and late 2005. Comparing the bank's average assets to its expected average assets reinforces this analysis.

Bank of America Admits Accounting Errors Related to Repo Transactions

In March 2010, John Hempton showed that Bank of America's average quarterly assets were consistently higher than its end of quarter assets. He argued this meant Bank of America was using Repo 105 style transactions to shrink its balance sheet before end of quarter reporting.

Only a few months later, Bank of America admitted "making six transactions that incorrectly hid from view billions of dollars of debt," according to a July 2010 Wall Street Journal article. As the Journal puts it, "BofA…erroneously classified some short-term repos as sales when they should have been classified as borrowings…" The article noted how "Bank of America's accounting of the six trades [was] similar to what…Lehman Brothers Holdings Inc. did to make its balance sheet look better before it filed for bankruptcy in 2008."

Interestingly, though Bank of America admitted doing this, the company also "said its incorrect accounting for the six trades wasn't intentional." The company said it "[does] not deliberately structure transactions that are economically disadvantageous simply for the purpose of…reducing recorded liabilities." In other words, Bank of America denied intentionally "parking its assets off balance sheet," as Hempton put it in his blog post.

Bank of America's admission also only covered the years 2007 to 2009. John Hempton used data from 2006 to argue that Bank of America was intentionally moving assets off-balance sheet to improve its balance sheet metrics. However, he also said in his post that "[it] was not just 2006 either - this had been happening for a while."

Evidence can be found for this. Using Hempton's method of comparing end of quarter assets to average assets, one can see similar results for 2005:

Source: Bank of America 2005 Annual Report

Thus, Bank of America's admission may have vindicated Hempton's analysis of the company's financial statements. However, that admission did not cover the entire timespan Hempton described. Moreover, the admission covered a much smaller dollar amount of accounting issues as well. Bank of America admitted to six mistaken transactions totaling $10.7 billion over three years, while Hempton's blog post argued that the bank was doing "$50 billion or more of overnight repos" each quarter. This is a difference of over an order of magnitude.

Nevertheless, by the end of 2010, it seemed as if Repo 105 style transactions were no longer an issue for Bank of America. The company had admitted at least some errors. In its 2010 letters to the SEC, it said that it hadn't made any of the relevant erroneously accounted-for trades since early 2009. According to the Journal, Bank of America also said that it had "strengthened its controls and re-emphasized to its staff the need to 'escalate' consideration of any unusual balance-sheet changes, in particular if they result from 'non-normal' transactions near the end of a quarter."

I Find Possible Continued Use of Repo 105 Style Transactions at Bank of America, 2010~2011

I first read John Hempton's blog posts about Repo 105 in late 2011. By that point, it seemed like such issues were little more than history. After all, his examples were from 2006. He had compared Bank of America's actions to those of Lehman Brothers before the crisis. Thus, I assumed that discrepancies between average and end of quarter assets were yet another of the issues that banks had resolved after the crisis.

Instead, when I looked at Bank of America's 2010 and early 2011 financial statements, this was what I found:

Source: Bank of America 2011 Annual Report

As you can see, in every quarter, Bank of America's average assets were still significantly higher than its end of quarter assets. My initial thought was that this might be because the bank was shrinking after the financial crisis. After all, as I mentioned earlier, a bank with shrinking assets should have higher average assets relative to end of quarter assets in each quarter.

However, you can account for that by considering the bank's expected average assets as well. If you do so, Bank of America's average assets still look too high:

Source: Bank of America 2011 and 2010 Annual Reports

I sent John Hempton my discovery that Bank of America still had much higher average assets relative to its end of quarter assets. This was his theory for what might be happening:

What I think is going on is that there are some traders who are judged by asset use - probably in Merrill Lynch.

They want to use the assets but NOT on balance date - so my guess is that the traders themselves are doing REPO105 type transactions to thwart the bank systems.

This makes sense. According to the Journal, Bank of America said that the erroneously accounted-for transactions it admitted to occurred in a particular business unit and were intended "to reduce the specific business unit's balance sheet to meet [the bank's] internal quarter-end limits for balance sheet capacity." This is more or less the same explanation that Hempton provides for the discrepancy between Bank of America's end-of-quarter and average quarterly assets through 2011.

A Continuing Balance Sheet Quirk At Bank of America: 2012~2015

Having discovered the quirk of how Bank of America's quarterly average assets were still higher than its end of quarter assets in every quarter, I initially planned to write about it in late 2011. Unfortunately, events intervened and I didn't have the chance to do so.

Earlier this year, though, I finally had the opportunity to re-explore Bank of America's balance sheet. I was curious if there was still a discrepancy between the company's average and end of quarter assets. To study this, I looked at the company's balance sheets from late 2011 through 2015.

The results of my investigation were mixed. On the one hand, Bank of America's average quarterly assets were no longer always higher than its end of quarter assets:

Source: Bank of America Annual Reports and Q2 2015 Financial Press Release

On the other hand, in most quarters, the bank's average assets were still higher than both its end of quarter assets and expected average assets. This discrepancy can be seen here:

Source: Bank of America Annual Reports and Q2 2015 Earnings Press Release

Even when the company's average assets were lower than its end of quarter assets, they were only lower by $10 billion or less. In contrast, when they were higher, they were often higher by $30 billion or more. This effect is even more dramatic when you compare average assets to expected average assets. The company's average assets were lower than its expected average assets in only three quarters, and only by small amounts.

Thus, there apparently has been a persistent upward bias since the end of 2011 in Bank of America's average assets relative to its end of quarter and expected average assets.

It is difficult to know what might be causing this trend. As I mentioned earlier, John Hempton believed that Bank of America had higher average assets relative to end of quarter assets from 2010 to Q3 2011 because traders were evading internal asset limits. This might also explain the more recent upward bias in average assets relative to end of quarter and expected average assets.

In any case, once I discovered this, I looked at the other largest U.S. banks' balance sheets. I wanted to see if this upward bias existed for them as well.

The first bank I looked at was Wells Fargo:

Source: Wells Fargo Quarterly Filings and Q2 2015 Earnings Press Release

As you can see, Wells Fargo exhibits the exact opposite trend from Bank of America. Bank of America's average assets were higher than its end of quarter assets in most quarters. They were also higher than its expected average assets in almost every quarter. In contrast, Wells Fargo's average assets were lower than its end of quarter assets in almost every quarter. They were also lower than its expected average assets in most quarters:

Source: Wells Fargo Quarterly Filings

This was also true between the fourth quarter of 2009 and third quarter of 2011. That was the period when Bank of America's average assets were highest relative to its end of quarter and expected average assets. In that time period, Wells Fargo still had lower average assets relative to end of quarter and expected average assets in almost every quarter.

Source: Wells Fargo Quarterly Filings

This creates some interesting conclusions. Since 2010, in most quarters, Bank of America has been riskier than it has appeared based on its leverage ratio. That is because the leverage ratio is calculated using end of quarter numbers, and its average assets have been higher in most quarters than at the end of the quarter. This means its average leverage is generally higher than its reported leverage.

In contrast, Wells Fargo has been safer than it has appeared based on its leverage ratio. This is because its average assets have been lower in most quarters than at the end of the quarter. This means that its average leverage has been generally lower than its reported leverage. Interestingly, this is not only the result of the company's asset growth. This can be seen by looking at the company's expected average assets in each quarter. Even relative to expected average assets, the company still shows a downward bias in its average assets in each quarter.

Interestingly, Wells Fargo is not the only bank to show this downward bias. JPMorgan Chase does as well:

Source: JPMorgan Chase Earnings Releases

Again, this is clear if you look at the difference between average and end of quarter assets for the company. The company's average versus expected average assets for each quarter are also similar to Wells Fargo's:

Source: JPMorgan Chase Earnings Releases

Source: JPMorgan Chase Earnings Releases

Again, as with Wells Fargo, JPMorgan Chase has been safer than it has appeared on the basis of its leverage ratio.

Like with Bank of America, it is hard to tell what might be causing this downward bias in average quarterly assets relative to end of quarter assets at JPMorgan Chase and Wells Fargo. In his blog post, John Hempton described how higher average assets relative to end of quarter assets implied that a bank was engaging in Repo 105 type transactions to shrink its end of quarter balance sheet. He also indicated how the reverse-lower average assets relative to end of quarter assets might mean that a bank is taking the other side of such transactions.

As he put it, Mitsubishi UFJ Financial Group (NYSE:MTU) had persistently lower average assets relative to end of quarter assets. In his words, "the differences [between end of period and average assets] and timing roughly match" those for Bank of America, but in reverse. As a result, he argued that MUFJ was Bank of America's counterparty in its 2006 Repo 105 style transactions.

Of course, that's not to say that JPMorgan Chase and Wells Fargo are operating as counterparties to other banks' Repo 105 style transactions. Rather, my only point is that this downward bias is potentially just as interesting as the upward bias in average assets at Bank of America. Unfortunately, it's beyond the scope of this article to consider what might be causing this downward bias.

Finally, it is worth considering the last of the four largest U.S. banks, Citigroup (NYSE:C):

Source: Citigroup Earnings Releases and 10-Q Filings

Citigroup's situation is interesting. In 2010 and 2011, the company's average quarterly assets were generally higher than its end of quarter assets. Similarly, the company's average assets were higher than its expected average assets.

However, the difference wasn't as drastic as with Bank of America. The difference between average and end of quarter assets was, on average, $36 billion in each quarter of 2010 and $24 billion in each quarter of 2011. In contrast, it was $111 billion and $76 billion, respectively, in those same time periods for Bank of America. This wasn't because Bank of America was that much larger. Citigroup had 80% of Bank of America's assets in 2010.

Moreover, Citigroup had one quarter in which its average assets were less than its end of quarter assets. In contrast, in every quarter, Bank of America's average assets were significantly higher than its end of quarter assets. Nevertheless, the upward bias in Citigroup's average assets was distinctive in this time period:

Source: Citigroup Earnings Releases and 10-Q Filings

In contrast, this upward bias disappeared in 2012 and has remained gone through the beginning of 2015:

Source: Citigroup Earnings Releases and 10-Q Filings

In some quarters, the company still has elevated average assets relative to end of quarter assets and expected average assets. However, that is offset by other quarters in which average assets are depressed relative to end of quarter and expected average assets. The mean difference since the end of 2011 between its average quarterly assets and end of quarter assets is a negligible $4 billion. The mean difference between its average assets and expected average assets is even smaller, at only $2.7 billion.

Thus, Citigroup had the same upward bias in its average assets in 2010 and 2011. However, this trend appears to have ended in 2012. These days, there does not appear to be an upward or downward bias in its average quarterly assets, relative to end of quarter and expected average assets.

Conclusions

In every quarter of 2010 and 2011, Bank of America had higher average assets relative to end of quarter assets. John Hempton attributed this phenomenon to traders evading internal controls. Moreover, in his blog post, he described how this sort of thing occurred at Bank of America before the financial crisis as well. He compared it to the Repo 105 transactions used by Lehman Brothers to alter its balance sheet in the quarters before it went bankrupt.

What I find interesting is that Bank of America's balance sheet still shows this same quirk of upwardly biased average assets. On average, the company's average assets in each quarter are higher than its end of quarter assets. Since the beginning of 2012, the mean difference has been $14 billion. That said, this is down significantly from $76 billion in 2011 and $111 billion in 2010.

Moreover, it is interesting that this quirk cannot be explained by normal balance sheet changes. As I mentioned, a bank with a shrinking balance sheet like Bank of America post financial crisis should have higher average assets in a quarter relative to end of quarter assets. I compensated for this by calculating the company's expected average assets for each quarter. However, even having done so, Bank of America's average quarterly assets since 2012 have appeared anomalously high.

Of course, this phenomenon wouldn't necessarily be strange if it existed at all of the banks. After all, there might be seasonal factors that cause banks' assets to be higher in the middle of a quarter than at the end of the quarter. However, looking at the other largest U.S. banks, it is clear that Bank of America is the only one of the "Big Four" banks that displays this behavior. Citigroup showed the same behavior, to a lesser extent, in 2010 and 2011, but does not do so anymore. JPMorgan Chase and Wells Fargo, oddly enough, actually show the opposite phenomenon. They actually have lower average assets each quarter than end of quarter assets. Thus, it seems clear that there's something different about Bank of America.

Final Thoughts: Does This Matter?

Of course, the question to ask is how all of this matters to investors. I recently mentioned to another investor that I was writing an article about an accounting quirk at Bank of America. He noted that there was a problem with finding such quirks at large companies like Bank of America.

As he put it, such quirks, though interesting, are often not material to an investing thesis in such companies. That is because such companies are so huge that even multibillion dollar quirks like the one I have described in this article don't really change the mix of information about the company.

Moreover, there is another issue with this particular quirk. On the one hand, it seems unlikely that the upward bias in Bank of America's quarterly average assets relative to end of quarter and expected average assets is just random noise. After all, none of the other "Big Four" banks exhibits this behavior. More importantly, it's similar to a symptom of past behavior which the company admitted was intended to evade internal controls.

However, there's no definitive way to "prove" that there's anything anomalous going on. The mean difference between average assets and end of quarter assets in each quarter between the beginning of 2012 and the second quarter of 2015 is only $14 billion. The standard deviation of this difference over the same time period is $21 billion. Thus, the difference isn't even one standard deviation from zero. If these results had come from a statistical sample, we would attribute this difference to random noise.

And yet, it's worth noting that this situation is different from a random sample. Rather than drawing data points from an overall population, we are working with the entire population of results. Moreover, we can look at these results through the filter of the company's past actions. After all, the mean difference between average and end of quarter assets in each quarter of 2011 was almost 10 standard deviations from zero. That is statistically significant, though admittedly based on an even smaller population size. Of course, that doesn't even take into account the actions of the company's competitors, which is significant in their own way. Because of this, it seems quite possible that something significant has been going on, even if it can't be shown by a statistical test.

In any case, if this quirk is significant, it offers some interesting potential lessons. First, it shows how incredibly complicated companies can be. As I've noted in several of my recent articles, as an outside investor, it is almost impossible to fully understand a company. The interplay between large numbers of employees, assets, and customers means that potentially significant quirks like this one are probably buried in every major company.

Beyond that though, this quirk may also offer a lesson about trends within Bank of America. The Journal described how Bank of America admitted that employees in one of its units made erroneous transactions to bring its balance sheet in compliance with internal limits. Such transactions resulted in lower end of quarter assets and thus an upward bias in average quarterly assets relative to end of quarter assets. The upward bias that can be seen in Bank of America's results since 2010 may have been caused by similar employee actions.

If it was, it may be significant that this upward bias has declined since 2012 and seems to have vanished in the first two quarters of 2015. This may be a sign of how Bank of America's internal controls have improved over that time period, preventing employees from affecting its balance sheet in the way they used to. If so, that sort of tangible evidence of improving internal controls would, in my mind, be valuable to investors in the company.

Thus, I will leave it up to my readers to decide whether this balance sheet quirk at Bank of America is significant or not. Moreover, I will leave it up to them to decide if this information affects their opinion of Bank of America as a potential investment. I can only say that when I learned about this quirk in late 2011, I thought it was interesting, but I still felt that Bank of America was a good investment.

Disclaimer: The content here is not meant as investment advice. Do not rely on it in making an investment decision. Do your own research. The content here reflects only the author's opinions. Those opinions might be wrong. This content is meant solely for the entertainment of the reader and its author.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

About this article:

Expand
Author payment: $35 + $0.01/page view. Authors of PRO articles receive a minimum guaranteed payment of $150-500.
Tagged: , Regional - Mid-Atlantic Banks,
Want to share your opinion on this article? Add a comment.
Disagree with this article? .
To report a factual error in this article, click here