Forbes Magazine's America's Best Banks: Are they best for investors?
Abstract
This paper investigates announcement effects and longer-term performance associated with the Forbes' America's Best Banks survey release. Although the market reacted positively to the announcement, the overall event window effects were insignificant. Raw and risk-adjusted returns are statistically insignificant over longer-time horizons. Investors cannot just use the Best Banks list to earn positive announcement window returns. However, a direct relation exists between movement in survey rank and subsequent accounting profitability measures, suggesting investors may benefit from monitoring movements on the Best Banks list. We also find support for a size effect as smaller, matched sample banks have higher Jensen's alphas than the Forbes larger Best Banks.
1 INTRODUCTION
Banks are complex organisations with many different types of assets, liabilities and off-balance sheet (OBS) items. These investments, funding for the investments and risk management tools used to protect the investments are challenging to understand for the most sophisticated investors from the perspectives of both performance and risk. Banking is heavily regulated, and the regulatory scrutiny and authority seem to wax and wane with the weaknesses and strengths of the industry and the overall economy.
While information on individual banks is available on websites such as the Federal Deposit Insurance Corporation (FDIC) site, it can be difficult for lay investors to access. Bank financial statements can also be challenging to interpret. In this study, we rely on rankings of an outside, non-regulatory source, Forbes Magazine, which examines and then ranks the performance of the US's largest financial institutions. We examine event window effects and the Forbes Best Banks' short- and long-term returns to determine whether investors can use the list to improve bank portfolio performance. We also examine whether movement on the list, either several positions up or down, is relevant to subsequent accounting performance.
For many reasons, investors may rely on an outside, independent party to better understand individual banks' relative risks and performance. Organisations such as Forbes and the American Bankers Association's ABA Banking Journal can improve transparency, which in turn bolsters investor confidence. Trust and credibility are also enhanced. Outside verification can help investors make better risk assessments of potential investments. Whether intentional or otherwise, banks may offer the public incomplete, selective or biased information regarding potential risks that impact future performance. In this study, we examine whether investors benefit by using the information provided by Forbes to make investment decisions.
Banks are multifaceted, and the larger the bank, the more complicated it is. In spring 2023, the Federal Reserve (2023) admitted in a report on the Silicon Valley Bank (SVB) failure that it ‘did not fully appreciate the extent of the vulnerabilities as SVB grew in size and complexity’. If one of the most powerful bodies in the world, the US Federal Reserve, admits to missing the risks and vulnerabilities of one of the country's largest banks, it seems clear that bank stock investors need reliable sources to help process information about the industry and the banks within it. We hypothesise that investors should use and benefit from information provided by various sources to identify the banking industry's best investment opportunities. Forbes Magazine's America's Best Banks list is a strong candidate for such information.
America's Best Banks list, published by Forbes Magazine, has been published since 2009, emerging from the financial crisis of 2007–2008. In late 2008, the US government took an unprecedented step and purchased stock in nine of the largest US banks. It was a partial nationalisation of the banking system and a unique time in the history of the United States and the industry. With dramatic changes impacting the banking industry, Forbes seized an opportunity to provide information to the public in unprecedented times. They began examining the performance of the US's largest banks. This information is useful to investors considering purchasing bank stocks and customers wanting to ‘invest’ via deposits in high-performing, safe banks. Since 2009, the company has continued publishing the best banks list, the most recent in February 2023.
Forbes uses bank profitability, credit quality and growth data of the 100 largest publicly traded banks and thrift institutions to rank the best to worst. Bank size is based on total assets. The four largest banks, JP Morgan Chase, Bank of America, Wells Fargo and Citigroup, hold $11.1 trillion of the $26.4 trillion held by the US banking industry. The 100 largest banks hold over 50% of banking assets (Forbes, 2023a, 2023b). However, as Bisnoff (2022) points out, smaller institutions dominate the top of the Forbes list, with only two banks with more than $100 billion in assets in the top 40 in the 2022 Best Banks list. Investors are less likely to be aware of the financial health of these smaller banks relative to the big four, which are routinely in the news. Also, as the spring 2023 bank failures show, it is even more critical to understand which of the banks are the most robust and profitable, as something as straightforward as poor interest rate risk management can cause the demise of a large commercial bank. Another advantage of surveys like the Forbes Best Banks list is that it makes government data on all banks, large or small, available and easy to understand.
FDIC call report data shares information similar to Forbes and is available to the public. The FDIC collects Reports of Condition and Income data on all insured national and state-chartered member banks each quarter. FDIC data has historically been challenging to access, hard to download efficiently and unwieldy. Unfortunately, recent changes by the FDIC have not improved usability or access much. The data is difficult to analyse in the aggregate. So, while some may argue that investors should go straight to the source and access the FDIC call report data themselves, that feat is impractical for many bank stock investors who do not have the technological or financial savvy to access and interpret the financial information of commercial banks. In other words, even if one can figure out how to access it now, the raw data is not particularly useful to many investors.
There is a broad literature regarding factors that affect individual investment decisions. Many studies are focused on behavioural finance (see Kumar & Goyal, 2015; Divanoglu & Bagci, 2018; Choi & Robertson, 2020, for example). Some older studies focus on investors' use of information analysts provide (see Hirst et al., 1995, for example). However, few studies directly address whether modern-day investors can benefit from the vast amount of information publishers such as Forbes Magazine provide. This study extends the literature on investor decision-making to include information provided by widely dispersed and read publications such as Forbes.
This study examines whether superior performance reported in the Forbes Best 100 Banks rankings translates into higher investor returns. We investigate both announcement effects surrounding the annual survey's release and survey-to-survey holding period returns on a raw and risk-adjusted basis. Contrary to our expectations, the overall event window period (−2, +2) CAR is 0.01%, which is insignificantly different from zero despite positive announcement day returns. This finding implies that an investor cannot simply use the Best Banks list to earn positive announcement window returns compared to investors who do not use the list to enhance investment decision-making.
On a longer-term basis, banks listed in the survey do not exhibit superior performance on a raw or risk-adjusted basis. However, an interesting result from these findings is that the matched sample Jensen's alpha is significantly positive (5% level) while the Forbes Best Banks Jensen's alpha is slightly positive but insignificant. We interpret this as evidence of a size effect in banking. The matched sample banks are, by definition, smaller in asset size. We inadvertently support a size effect in bank stocks because of how Forbes constructs its Best Banks list (i.e., the 100 best but largest banks). We also find a direct relationship between survey rank movement and subsequent accounting profitability measures. For banks that move up (down) the list substantively, accounting performance is either better (or worse) the following year. This finding implies some value to investors, even though it is not, as we expected, higher returns than a matched sample simply by investing in the Forbes listed banks.
1.1 Literature review
In this study, we examine if higher shareholder returns result from banks lauded by Forbes based on solid accounting measures. Previous work has looked at other data sources with a similar mission. Filbeck et al. (2013, 2016) use the American Bankers Association ABA Banking Journal to identify the strongest performing banks. The first of the ABA's annual surveys appeared in 1993 with rank based on return on assets (ROA), shifting in 1999 to return on equity (ROE). Filbeck et al. (2013) examine banks listed on the ABA surveys from 1993 to 2009. They look at the entire sample period and break it into sub-periods (e.g., 1993–1998, 1999–2000 and 2001–2009). The sub-periods are based on how the ABA defines ‘best performers’ in the various periods, moving from ROA as the relevant metric in the early years to ROE in the latter years. The authors investigate whether the ABA's Banking Journal Top Performing Banks survey results in higher investor returns. They find that for the entire survey period and later subperiods based on ROE, the top-performing banks outperform the S&P 500 index on a raw and risk-adjusted basis, including buy-and-hold abnormal returns (BHARs) compared with its matched banks. These results remain robust after controlling for market return, capitalisation, style and momentum factors.
Considering the best and worst banks through a different lens, Filbeck et al. (2016) examine bank performance around the financial crisis of 2007–2008. In keeping with Peters and Waterman's In Search of Excellence: Lessons from America's Best Run Corporations (Peters & Waterman, 1982), the 2016 Filbeck et al. study explores whether banks with the best historical, accounting-based performance pre-financial crisis produce higher shareholder returns post-financial crisis. The authors examine the worst and best-performing banks on the ABA list of best and worst performers around the financial crisis of 2007–2008. Filbeck et al. specifically consider performance around the crisis compared to peers and other benchmarks. They also examine the use of leverage by banks leading up to the crisis as an indicator of increased risk, which could result in inferior performance or bank failures during and following the crisis, a time of severe financial distress. While Filbeck et al. do not find support that the worst-performing banks on the ABA's list were less leveraged (and thus exhibited less financial risk) leading up to the crisis, they do find that following the 2007–2008 financial crisis, the worst-performing banks sample outperformed the best-performing banks sample on a raw and risk-adjusted basis.
In this study, we update and extend the work of Filbeck et al. (2013, 2016) by investigating a different and more broadly read source of the identification of ‘best banks’, Forbes Magazine. We examine whether the Forbes Best Banks outperform peers and the market over both the event window and a longer investment horizon.
The relationship between a company's size and performance has interested researchers for many years. Gandhi and Lustig (2015) examine the relationship between bank size and risk-adjusted returns. There are many studies on the relationship between size and return (e.g., Banz, 1981, Fama & French, 1993, and many others) of non-financial firms. These studies avoid financial firms due to their high leverage ratios (e.g., commercial banks typically have 10 percent or less equity to total assets). Gandhi and Lustig are the first to document a size effect in financial firms and show that performance is about size, not market capitalisation. The authors find that the largest banks earn lower risk-adjusted returns than small to mid-sized banks, attributing it to how tail risks are priced in financial stocks. The Forbes list of Best Banks begins with the 100 largest banks and then ranks them on several performance, growth and risk measures. This methodology means that bank size, not just bank performance, is relevant to our study. We are examining the relative performance of the 100 largest banks, not the literal best and worst banks in the universe of banks.
In the post-financial crisis period, capital requirements have increased for banks, but most significantly for the largest or systemically important financial institutions (SIFIs). Berger and Bouwman (2013) examine the impact of bank capital on two measures of performance, survival and market share, in the wake of the 2007–2008 financial crisis. The authors examine bank performance and capital levels in both times of severe stress, in the market and banking industry, and in normal economic periods. They find that better-capitalised small banks are more likely to survive in all economic scenarios, market crises, financial industry crises and normal periods. For medium and large banks, more capital does not seem to influence survival in normal periods or periods of market stress but does enhance performance during banking crises.
In a related study, Bouwman et al. (2018) examine bank stock price performance and capital. In contrast to the commonly held belief that higher capital drags down performance, the authors find that well-capitalised banks do not perform worse than the more heavily leveraged banks. The authors find that high-capital bank stocks earn higher alphas (i.e., risk-adjusted returns) than low-capital banks, both in and out of the sample. They find the results robust to different asset pricing models, capital definitions, crisis periods, and ex-ante expected returns.
Ariff et al. (2013) use different information to test bank performance – unexpected earnings announcements. The authors examine whether bank share prices change significantly as non-bank share prices do ahead of accounting earnings reports. They find that, like non-financial firms, share prices in European banks from eight countries respond to unexpected earnings announcements at the time of accounting reports. The authors examine only the top 10 to 12 banks in each country and conclude that banks respond similarly to non-bank stocks to unexpected earnings. However, they concede that the sample may be biased as only eight of 18 countries are included.
Researchers have considered other factors that affect investors' buying and selling behaviour. In an earlier study, Docking et al. (1997) examine the announcement effects of loan loss reserve (LLR) announcements. Unlike the ‘Best Banks’ announcement by Forbes or the ABA's Banking Journal, this study examines an announcement that conveys potentially negative news as banks add to their LLR to cover possible loan write-offs or loan portfolio revaluations. The authors contend that the signal is important to investors as higher LLRs can imply riskier loan portfolios or signal the potential for lower earnings and dividend payments. The announcement of an increase in the loan loss reserve can signal important asymmetric information regarding credit conditions in the bank. The authors find negative announcement effects on LLR announcements. The negative response is greater for regional (i.e., smaller) banks than for money centre banks. However, they find that the announcement effect is nullified when the LLR news is paired with favourable news, such as positive earnings announcements. Suppose the LLR announcement is instead coupled with ‘bad news’, such as lower expected earnings or announcements of dividend reductions. Investors are most threatened in those cases, resulting in significant negative stock price effects. Thus, not all bank announcements result in positive abnormal returns for investors.
What constitutes the best measure of bank performance? One would expect bank stock analysts to incorporate historic accounting-based metrics such as earnings per share (EPS), ROA and ROE as one method by which to evaluate the merits of a bank or bank holding company (BHC) stock when making buy, sell or hold recommendations. However, investors consider various factors regarding which securities to buy and sell. In past decades, analysts were a primary source of investors' information (Han & Wild, 1991; Pownall et al., 1993). However, investors increasingly respond to other ‘news’ sources based on behavioural or ‘noise’ factors. Take the GameStop situation in February 2021. Analysts were not touting the stock, yet investors were paying attention to the ‘news’ they heard and read. Reddit's WallStreetBets group, other social media sites and everyone else, from the stodgy nightly news to friends on Snapchat, were discussing the stock as a ‘should you or shouldn't you jump in?’ Statman (1999) and other researchers have considered alternative frameworks by which investors approach investment decision-making. He points out that some investors are prone to representativeness, employing strategies that chase past winners or associate good or well-run companies with suitable investments.
Cates (1996) referred to this period as the ‘Ice Age’, a time when investors relied heavily on analysts' recommendations and banks primarily generated revenues from securities and loans funded by deposits (i.e., from the balance sheet). He argues that ROA was an appropriate performance measure since revenues and expenses originated on the balance sheet. However, banking has changed dramatically and continues to evolve. Bank profits and risks are increasingly associated with off-balance-sheet activities, especially for large banks. This trend means total assets, the denominator of ROA, is no longer the sole driver of net income, the numerator of both ROA and ROE. Cates suggests ROE is the more meaningful performance measure for modern banks because shareholders underwrite both banks' on- and off-balance sheet risks.
Pennacchi and Santos (2021) examine the topic of ROE as a primary target performance measure of banks in recent years. Like most non-financial firms, the authors note that banks targeted EPS as the industry's primary performance metric in the 1970s. While non-financial firms still target EPS, banks have recently shifted their emphasis and efforts toward ROE. According to the authors, one key issue that drove banks to focus on ROE was flat-rate deposit insurance premiums. Regardless of risk, all banks paid a flat rate to the FDIC until 1993. In 1993, the system changed to a risk-based deposit insurance premium system. However, due to statutory restrictions, the FDIC was severely limited regarding who it could charge. Despite depositors being insured due to regulations, well-capitalised and highly-rated banks did not pay for FDIC insurance. Thus, 5 percent of institutions paid deposit insurance premiums under the new risk-based system. In 2006, Congress changed the law so that all banks had to pay FDIC insurance premiums. But Pennacchi and Santos argue that fixed-rate deposit insurance premiums and increasing competition in the banking industry pushed banks to focus on ROE rather than EPS as the key target performance metric. ROE reflects EPS growth and shows managers' efficiency in generating profits. Investors can glean insights into margins, revenue and retained earnings from ROE. Forbes uses several metrics of accounting performance, including return on average equity, return on average assets, the efficiency ratio, among others. We agree that, in general, ROE is a better performance metric than ROA for a modern commercial bank.
As noted, Forbes examines several variables to determine the best-performing banks. Banks adopted balance sheet return measures as performance measures in the 1960s. We examine several return measures in this study, using the Forbes metrics/Best Banks list as the jumping-off point for analysis of abnormal returns, buy and hold returns and other measures.
1.2 The Forbes Best Banks
The investment landscape is changing, and new investors seem open to a wider variety of information sources about stocks rather than relying solely on the advice of brokers and financial analysts. Authors from the Financial Industry Regulatory Authority (FINRA) Investor Education Foundation studied investors entering the market in 2020 and the factors influencing their decision-making (Lush et al., 2021). The study focuses on survey results. The authors find that new entrants into the investment landscape in 2020 were younger, had lower incomes and were ethnically diverse. Saving for retirement was cited as the primary reason for opening an investment account, closely followed by the ability to invest small sums of money, which online trading platforms allow. Even this new mode of trading via online platforms puts distance between the investor and the broker. Younger investors do it independently, without a broker giving buy and sell advice based on the recommendations of the firm's analysts. According to the study, new investors were unlikely to rely on traditional sources of information, such as financial analysts, to make investment decisions. This pattern means they need accessible, alternative sources of information. Forbes Magazine offers that accessible alternative.
Based on our analysis, the Forbes Best Banks list's value to the investing public is of interest. Is this a source of information that the new generation of investors might use to generate alpha? This study is similar to other research investigating whether a group of admirable/well-managed/top-performing firms are also investor-worthy. Clayman (1987) was one of the first to examine the performance of admirable firms in the book In Search of Excellence. Researchers have followed Clayman, attempting to determine which surveys and lists are worthy of investors' attention.
We contribute to the literature in this study, examining whether investors who use the Forbes America's Best Banks survey can generate enhanced returns. Forbes is a long-standing brand, publishing its first issue in 1917. According to the Forbes website, the publication has over six million readers. The ‘Best Banks’ annual articles appear high (in the top two or three websites) in search results if you search topics such as ‘best US banks’ or ‘top bank performers’, potentially increasing readership further to include non-subscribers and those interested in bank stocks. This high level of readership distinguishes Forbes from other publications that examine bank performance, such as the ABA Banking Journal, which, according to the publication, has an audited circulation of 34,584. Additionally, the Banking Journal targets bankers, regulators and others in the industry. It is, in essence, a trade journal. This study furthers the work of Filbeck et al. (2013) by examining the performance of financial institutions brought to light by a widely read publication that has a greater likelihood of reaching a broader audience and one that is not necessarily focused on industry insiders.
The Forbes 2023 Media Kit (Forbes, 2023a, 2023b) indicates that Forbes has a digital and social media reach of 140 million readers, with 1 billion YouTube views in 2021 and over 50 million social footprints. In addition, Forbes reaches 5 million print readers and is ranked first for millennials. Forbes appears in 49 global editions in 83 countries and is available in 28 languages. They note that 93 percent of high-net-worth individuals say Forbes helps shape their opinions on markets, investing and economic outlooks. Ross (2021) states that Forbes is one of the six most influential and dominating global business magazines that have helped shape opinions and ideologies. As Forbes is highly read and influential, we expect investors to use this information in decision-making. We examine whether investors are ‘paying attention’ by examining the cumulative abnormal returns (CARs) around the announcement of the Best Banks. We also examine whether investing accordingly can lead to higher, long-term, raw and risk-adjusted returns.
The metrics considered to make this list include: non-performing assets as a percentage of total assets; reserves as a percentage of non-performing assets; CET1 ratio, which compares a bank's capital against its risk-weighted assets; risk-based capital ratio; return on average tangible common equity; return on average assets; net interest margin; operating revenue growth; and net charge-offs as a percentage of total loans. This list excluded banks where the top-level parent company is not based in the U.S.
While investors may go directly to the banks' financial statements on the FDIC's website, the information is difficult to parse. Additionally, Forbes aggregates different types of accounting information, from capitalisation to return measures to non-performing loans to rank banks. At least from the perspective of a typical retail investor, this aggregation, and more importantly, understanding the relevance of each type of information to the overall health and performance of the institution, would be challenging at best.
Many investors could use the raw data if they were looking solely at the ROAs and ROEs of banks and rank-ordering them. But understanding the role of capital, the relative risk-adjusted capital levels, loan charge-offs over the year, the efficiency of banks managing non-interest expense, etc., requires more complex analysis and investment analysis skills than many investors possess. The advantage of the Forbes list accrues more to retail investors than institutions with greater access and expertise.
This point parallels the one made by Filbeck et al. (2013). They also note that the data in the ABA surveys are available to the public from 2001 (the survey began in 1993) but in a relatively inaccessible form – FDIC call reports. If banking data were easily accessible, the announcement effects of rankings based on publicly available information might not be expected to produce either short- or long-term effects. This lack of reaction is likely true given the greater transparency of banks' financial conditions due to regulation. However, in the case of the ABA surveys, Filbeck et al. (2013) do find the rankings of top-performing banks, as identified by the ABA's Banking Journal, outperform the S&P 500 index, both on a raw and risk-adjusted basis and buy and hold abnormal returns (BHARs) basis. Thus, one possible explanation is the market finds value in the release of the ABA survey due to the inaccessibility of the data. Can the same be said about the Forbes Best Banks? If so, this supplementary data might prove advantageous to investors.
This study links the book value return measures reported in the Forbes lists to shareholder returns. The advice and influence of friends, family and media sources inspire some investors. This study investigates whether investors pay attention to the Forbes list of Best Banks, one method investors might employ to identify ‘good or well-run’ financial institutions for investment. Based on the work of Filbeck et al. (2013), we hypothesise that the market will react positively to being named on the Forbes list. We also expect the Best Banks to outperform a matched sample, exhibiting superior long-term, risk-adjusted performance. Based on previous research regarding the size effect on financial institutions (e.g., Gandhi & Lustig, 2015), we hypothesise that relatively smaller banks on the list will outperform larger ones. We also examine banks' subsequent performance following substantive moves up or down the Best Banks list and expect that a large move up or down the list will result in better (or worse) accounting performance in subsequent periods. This paper's focus and contribution is to determine if an ex-ante investment strategy will ‘work’ based on investing in the Forbes Best Banks.
2 SAMPLE
- The sample firms must have return records on the Center for Research on Stock Prices (CRSP) Daily Combined Return File, 301 trading days (i.e., one trading year) immediately before the announcement date.
- The sample firms must have return records on the CRSP Daily Combined Return File after the announcement date until the following survey's press release date.
- The firm must have complete data on S&P's Research Insight® (Compustat Bank Annual file2).
The number of observations related to banks named in the Forbes survey between 2011 and 2022 is 1200. Across the 12 years of the survey, 1037 listings, comprising our whole sample, meet the selection criteria. We then further divide the entire sample into Best Banks, Worst Banks and Medium Banks samples according to their annual ranks: banks that belong to the top (bottom, medium) and third ranks belong to the Best (Worst, Medium) Bank sample. Further, we develop two subsamples to explore whether ranking changes affect banks' stock and accounting performance: Improve Rank sample and Deteriorate Rank sample. Improve (Deteriorate) Rank samples include banks whose ranking is improved (worsened) by at least 203 places to the previous year's ranking.
To evaluate the performance of the Forbes samples, we compare them to a matched portfolio that we use as a benchmark. Our matched portfolios are determined using propensity score matching (PSM) based on five of the ranking criteria of the Forbes listing (i.e., previous year-end return on equity (ROE), non-performing assets/total assets ratio, efficiency ratio, leverage and operating revenue growth rate). Previously ranked banks are excluded from being matched. We cannot match on size (i.e., total assets), as the jumping-off point for the Forbes list is the largest 100 banks in the industry. Therefore, our matched sample banks are smaller than the Forbes-listed banks. They are, however, matched on accounting performance measures (i.e., ROE), asset quality (i.e., non-performing assets to total assets), efficiency (i.e., managing overhead expenses), the use of leverage (i.e., total debt to total assets) and the growth in revenue, both interest and non-interest income. Our remaining pool of potential matched firms comes from the remaining banks with data available through the Compustat Bank Annual file.
Table 1 reports the summary statistics of the Whole sample and subsamples. Table 1 shows that the Best Bank sample has higher profitability ratios (i.e., ROE and ROA) than the Worst and Medium Bank samples.4 This relation is primarily caused by profitability being one of Forbes' ranking criteria. The Worst Banks sample has a larger total asset average than the Best5 and Medium Rank samples. For example, the Worst Banks sample's average total assets are $191.9 billion, while the corresponding number for the Best (Medium) Banks sample is $31.6 billion ($161.7 billion), respectively. Table 1 also shows that Improve Rank and Deteriorate Rank samples have smaller total assets than the whole sample.6 Our matched sample has similar profitability ratios (as one of the matching criteria is ROE) but much lower total assets than the Whole Forbes sample.7
| Variable | Number of bank year | Mean | Standard deviation | Percentile | ||||
|---|---|---|---|---|---|---|---|---|
| Min | 25 | 50 | 75 | Max | ||||
| Total assets ($M) | ||||||||
| Whole sample | 1037 | 129,934.49 | 436,238.69 | 628.88 | 12,457.25 | 21,275.65 | 46,651.55 | 3,743,567.00 |
| Matched whole sample | 1037 | 59,465.49 | 273,799.43 | 95.87 | 955.63 | 2054.15 | 4617.86 | 2,634,139.00 |
| Best Bank sample | 336 | 31,583.64 | 56,581.84 | 4722.83 | 11,795.44 | 17,753.48 | 26,635.38 | 444,438.00 |
| Worst Bank sample | 367 | 191,944.84 | 524,735.63 | 628.88 | 13,665.02 | 28,075.94 | 75,965.50 | 3,169,495.00 |
| Medium Bank sample | 333 | 161,691.76 | 521,629.87 | 691.38 | 11,969.60 | 20,973.71 | 66,112.01 | 3,743,567.00 |
| Improve Rank sample | 95 | 74,131.38 | 215,278.57 | 691.38 | 11,422.56 | 19,464.49 | 50,616.43 | 1,930,115.00 |
| Deteriorate Rank sample | 94 | 68,270.13 | 120,524.67 | 5547.04 | 13,300.20 | 24,028.25 | 44,683.66 | 573,284.00 |
| Previous year ROE (%) | ||||||||
| Whole sample | 1037 | 8.58 | 5.49 | −58.54 | 6.86 | 8.68 | 10.78 | 34.33 |
| Matched whole sample | 1037 | 8.41 | 6.03 | −64.82 | 5.61 | 8.44 | 11.06 | 52.91 |
| Best Bank sample | 336 | 9.60 | 4.09 | −34.43 | 8.00 | 9.61 | 11.22 | 19.74 |
| Worst Bank sample | 367 | 7.44 | 6.76 | −58.54 | 5.82 | 7.75 | 10.08 | 34.33 |
| Medium Bank sample | 333 | 8.77 | 4.96 | −56.33 | 7.29 | 8.48 | 10.62 | 27.16 |
| Improve Rank sample | 95 | 9.62 | 4.56 | −7.35 | 7.27 | 9.14 | 11.34 | 27.16 |
| Deteriorate Rank sample | 94 | 7.65 | 9.43 | −58.54 | 6.74 | 8.34 | 10.51 | 20.30 |
| Previous year ROA (%) | ||||||||
| Whole sample | 1037 | 1.31 | 0.65 | −5.60 | 1.09 | 1.35 | 1.61 | 2.88 |
| Matched whole sample | 1037 | 1.18 | 0.70 | −2.56 | 0.82 | 1.19 | 1.52 | 7.29 |
| Best Bank sample | 336 | 1.54 | 0.57 | −3.94 | 1.28 | 1.57 | 1.80 | 2.88 |
| Worst Bank sample | 367 | 1.07 | 0.73 | −5.60 | 0.89 | 1.21 | 1.43 | 2.27 |
| Medium Bank sample | 333 | 1.33 | 0.54 | −4.98 | 1.11 | 1.34 | 1.57 | 2.80 |
| Improve Rank sample | 95 | 1.39 | 0.51 | −0.40 | 1.07 | 1.41 | 1.71 | 2.71 |
| Deteriorate Rank sample | 94 | 1.16 | 0.98 | −5.60 | 1.08 | 1.32 | 1.51 | 2.46 |
- Note: Regarding the maximum number for the matched whole sample in this table, the value belongs to HSBC, which is surprisingly not listed in the Forbes top 100 banks list. HSBC Holdings PLC is headquartered in the UK but traded on the NYSE in USD.
3 METHODOLOGY AND RESULTS
3.1 Short-run market impacts
We hypothesise that there will be a positive market reaction to the release of the Forbes Best Banks list. Alternatively, the market may not respond to the announcement, which would indicate that either the market does not incrementally value the information in the survey or that the ‘news’ contained in the survey is already fully valued.
We assess the share price response to the survey's release, beginning two trading days before and ending 2 days after due to the possibility of news leakage or a delayed market acknowledgement of the survey's release. We calculate daily abnormal returns (ARs) and cumulative abnormal returns (CARs) over the event window of (−2, +2).8 Expected returns are estimated using the Fama–French 3-factor plus momentum model during the interval (−2, +2). Estimates of the parameters are calculated for the period (−326, −71). We follow Dodd and Warner (1983) and employ a standard event-study methodology.
Table 2 reports the event study results of the whole sample and subsamples. In the days leading up to the announcement, there was a negative market response over the event window (−2, −1). The whole sample and all subsamples show a statistically significant negative abnormal return over the event window (−2, −1) and a positive abnormal return over the event window (0, +2). All samples and subsamples show a statistically insignificant abnormal return over the event window (−2, +2).9
| Panel A: Abnormal returns (%) around the event date | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Day | Whole sample | Best Bank sample | Worst Bank sample | Medium Bank sample | Improve Rank sample | Deteriorate Rank sample | Whole Bank portfolio | |||||||
| AR | t-stat | AR | t-stat | AR | t-stat | AR | t-stat | AR | t-stat | AR | t-stat | AR | t-stat | |
| −2 | −0.23 | −6.45*** | −0.22 | −3.99*** | −0.35 | −4.55*** | −0.14 | −2.62*** | −0.34 | −2.81*** | −0.40 | −3.63*** | −0.12 | −4.15*** |
| −1 | −0.18 | −3.42*** | −0.20 | −2.27** | −0.25 | −2.54*** | −0.09 | −1.10 | −0.47 | −2.28** | −0.27 | −1.41 | −0.25 | −6.73*** |
| 0 | 0.38 | 8.87*** | 0.34 | 5.40*** | 0.49 | 5.94*** | 0.30 | 4.11*** | 0.45 | 2.73*** | 0.61 | 4.15*** | 0.20 | 6.01*** |
| 1 | −0.10 | −2.27*** | −0.13 | −2.01** | −0.10 | −1.16 | −0.06 | −0.99 | −0.14 | −0.91 | −0.25 | −1.69 | 0.06 | 1.91* |
| 2 | 0.14 | 2.78*** | 0.14 | 1.80* | 0.17 | 1.55 | 0.13 | 1.69* | 0.30 | 2.17** | 0.07 | 0.53 | 0.11 | 3.17*** |
| Panel B: Cumulative abnormal returns (%) around the event date | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Interval | CAR | t-stat | CAR | t-stat | CAR | t-stat | CAR | t-stat | CAR | t-stat | CAR | t-stat | CAR | t-stat |
| (−2, −1) | −0.42 | −6.05*** | −0.42 | −3.86*** | −0.61 | −4.37*** | −0.23 | −2.20** | −0.81 | −3.15*** | −0.67 | −2.81*** | −0.37 | −8.38*** |
| (0, +2) | 0.44 | 5.72*** | 0.36 | 3.31*** | 0.56 | 3.64*** | 0.38 | 3.06*** | 0.65 | 2.49** | 0.44 | 2.21** | 0.37 | 6.87*** |
| (−2, +2) | 0.01 | 0.05 | −0.07 | 0.45 | −0.06 | −0.28 | 0.13 | 0.37 | −0.20 | −0.64 | −0.24 | −0.97 | 0.65 | 0.10 |
- Note: This table reports the event study results for Forbes Best Banks. Panel A shows the abnormal returns (in percent) around the event date, and Panel B shows the cumulative abnormal returns (in percent) around the event date. We evaluate the share price response to the release of this survey beginning 5 days before the event date by calculating abnormal returns (ARs) and cumulative abnormal returns (CARs). Expected returns are estimated from the Fama–French 4-factor model. Expected returns are estimated during the interval (−2, 2), and estimates of the parameters are calculated for the trading day period (−301, −46) using 255 trading day years. We follow Dodd and Warner (1983) and employ an event-study methodology. *, ** and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.
To evaluate whether some confounding factors drive our event study results, we constructed a whole bank portfolio that comprises all banks with available return data and accounting data from the Compustat Bank Annual file. Our results for the Whole bank portfolio surrounding event dates in Table 2 show similar abnormal returns and CARs patterns compared with the Forbes Whole sample, suggesting no significant market reaction surrounding the announcements of the Forbes 100 largest banks listing.
3.2 Long-term stock return performance
Next, we examine the long-term return performance of the Whole and subsamples after each publication date. Barber and Lyon (1997), Fama (1998) and Loughran and Ritter (2000) have shown that the magnitude, and sometimes even the sign, of the long-run abnormal returns are sensitive to alternative measurement methodologies. We use a variety of methodologies to address this issue.
We compare each sample's raw monthly and risk-adjusted returns against the S&P 500 index. In addition, we calculate a variety of risk-adjusted performance measures. Then, we employ the Fama–French 3-factor and 5-factor models to assess the abnormal returns. Last, we compare the buy-and-hold returns of our Whole sample with the matched sample. The methodologies and test results are discussed in the following sections.
In Equation (1), represents the mean difference between the market and portfolio return each day, sd is the standard deviation of the daily difference between the returns of the market and portfolio, and n is the number of days corresponding to the annual holding period.
While the comparison of raw return data gives some information concerning each portfolio's performance, we gain little information regarding the level of risk contained in the portfolios. We calculate three commonly used risk-adjusted measures for comparison purposes.
In Equation (2), d represents the mean monthly difference between the Best and Worst Banks' portfolio and the T-bill return, calculated over respective holding periods, and sd is the sample standard deviation of the monthly return differences.
In Equation (3), d represents the mean daily difference between the return on the Best and Worst Banks portfolio and the T-bill return, calculated over respective holding periods, β is the portfolio beta, and n is the number of days in the respective holding periods. The Sharpe Index measures return per unit of total risk, and the Sharpe ranking is particularly relevant if an investor holds a poorly diversified portfolio. If an investor holds a diversified portfolio, Treynor's measure of return per unit of systematic risk is the more appropriate performance measure.
A positive (negative) alpha is consistent with an undervalued (overvalued) securities portfolio. Our null hypothesis is that there will be insignificant differences in the risk-adjusted measures of the Best and Worst Banks' portfolios compared to the S&P 500 index.
Table 3 reports the raw monthly returns and risk-adjusted measures for the samples. The cumulative raw returns show that the Whole sample and subsamples have higher raw returns than the S&P 500 index and whole bank portfolio. However, none are statistically significant, except for the Worst bank sample. The Whole sample and all subsamples have lower raw returns than the matched sample, but none are statistically significant, except for the Deteriorate Rank sample. The whole sample and all subsamples have lower Sharpe measures but higher Treynor measures than the S&P 500 index and lower Sharpe and Treynor measures than their matched samples. Jensen's alpha measures show that the whole sample and all subsamples have positive alphas, but none of these alphas are statistically significant. However, Jensen's alpha measures of the matched sample have statistically significant positive alphas except for the matched Improve Rank sample. We interpret this result as evidence of a size effect in banking as the matched sample banks are, by definition, smaller than the Forbes Best Banks.
| Whole Bank sample | Best Bank sample | Worst Bank sample | Medium Bank sample | Improve Rank sample | Deteriorate Rank sample | |
|---|---|---|---|---|---|---|
| Monthly raw return (%) | ||||||
| Test sample (1) | 1.211 | 1.110 | 1.331 | 1.166 | 1.282 | 1.201 |
| Matched sample (2) | 1.429 | 1.303 | 1.544 | 1.429 | 1.398 | 1.830 |
| S&P 500 Index (3) | 0.841 | 0.841 | 0.841 | 0.841 | 0.902 | 0.902 |
| Whole bank portfolio (4) | 0.958 | 0.958 | 0.958 | 0.958 | 1.142 | 1.142 |
| (1)–(2) | −0.243 | −0.215 | −0.239 | −0.290 | −0.156 | −0.725* |
| (1)–(3) | 0.370 | 0.269 | 0.491 | 0.325 | 0.380 | 0.299 |
| (1)–(4) | 0.253 | 0.152 | 0.373* | 0.207 | 0.139 | 0.058 |
| Sharpe measure | ||||||
| Test sample (1) | 0.032 | 0.029 | 0.034 | 0.032 | 0.034 | 0.030 |
| Matched sample (2) | 0.058 | 0.051 | 0.057 | 0.059 | 0.047 | 0.058 |
| S&P 500 Index (3) | 0.034 | 0.034 | 0.034 | 0.034 | 0.038 | 0.038 |
| Whole bank portfolio (4) | 0.029 | 0.029 | 0.029 | 0.029 | 0.036 | 0.036 |
| Treynor measure | ||||||
| Test sample (1) | 0.048 | 0.045 | 0.052 | 0.047 | 0.053 | 0.048 |
| Matched sample (2) | 0.092 | 0.084 | 0.093 | 0.097 | 0.086 | 0.107 |
| S&P 500 Index (3) | 0.039 | 0.039 | 0.039 | 0.039 | 0.041 | 0.041 |
| Whole bank portfolio (4) | 0.039 | 0.039 | 0.039 | 0.039 | 0.048 | 0.048 |
| Jensen's alpha | ||||||
| Test sample (1) | 0.012 | 0.008 | 0.016 | 0.010 | 0.013 | 0.008 |
| Matched sample (2) | 0.039** | 0.033** | 0.043** | 0.041*** | 0.034 | 0.053** |
| Whole bank portfolio (4) | 0.001 | 0.001 | 0.001 | 0.001 | 0.008 | 0.008 |
- Note: This table reports the raw and risk-adjusted returns of Forbes' Best Banks sample compared to the S&P 500, the bank portfolio and its matched sample. Panel A reports the monthly raw returns for the Forbes banks sample, the S&P 500 index, the bank portfolio and its matched sample. Panel B shows calculations for the three risk-adjusted performance measures: Sharpe, Treynor and Jensen's alpha. *, ** and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.
Overall, we find statistically insignificant performance of the Whole sample and all subsamples compared with the S&P 500 index on a raw return basis and using Jensen's alpha. The Forbes sample and subsamples have higher Treynor measures but lower Sharpe measures, the latter of which is not unexpected given the higher idiosyncratic risk of the banking portfolio. The Forbes sample and subsamples exhibit inferior performance compared with their matched samples. Again, we interpret this to have more to do with size than with the inferiority of the Best Banks. The matched sample banks are smaller than the Forbes-listed banks.
In Equations (5) and (6), Rpt represents the simple portfolio return on the Forbes bank sample, Rft is the return on one-month T-bills, Rmt is the return on a value-weighted market index, SMBt represents the return on a value-weighted portfolio of small stocks less the return on a value-weighted portfolio of big stocks, HMLt is the return on a valued-weighted portfolio of high book-to-market stocks less the return on a value-weighted portfolio of low book-to-market stocks, RMWt is the return on a value-weighted portfolio of prior robust operating profitability stocks less the return on a value-weighted portfolio of the weak operating profitability stocks, and CMAt is the return on a value-weighted portfolio of prior conservative investment stocks less the return on a value-weighted portfolio of the aggressive investment stocks.
A positive intercept for these regressions indicates that the sample firms have performed better after controlling for the market, size, book-to-market ratio, operating profitability and investment factors in returns. To determine whether the regression intercepts are significantly different from zero, we report the t-statistics from the regression intercept.
Table 4 presents the results of the Fama–French models. Our results indicate that after controlling for additional factors specified by the Fama–French models, we do not observe superior performance for the Whole and subsamples, as none of the intercepts, except for the Worst Bank Sample, are statistically significant. Since the Worst Bank sample is composed of larger banks (as noted in the discussion of Table 1), these findings are consistent with Gandhi and Lustig (2015), who find that the largest banks earn lower risk-adjusted returns than small to mid-sized banks. The Forbes listing consists of the 100 largest banks, so subsample characteristics are on a relative, not absolute, size comparative.
| Whole Bank sample | Best Bank sample | Worst Bank sample | Medium Bank sample | Improve Rank sample | Deteriorate Rank sample | ||
|---|---|---|---|---|---|---|---|
| Panel A: Fama–French 3-factor model | |||||||
| Intercept | Coefficient | 0.0072 | 0.0038 | 0.0109 | 0.0058 | 0.0036 | −0.0021 |
| t-stat | 0.63 | 0.3 | 0.93 | 0.52 | 0.24 | −0.15 | |
| MKTRF | Coefficient | 1.0601 | 1.0301 | 1.1049 | 1.0433 | 1.0606 | 1.0864 |
| t-stat | 104.22*** | 89.96*** | 104.46*** | 104.25*** | 76.87*** | 80.35*** | |
| SMB | Coefficient | 0.6835 | 0.7372 | 0.6716 | 0.6461 | 0.7241 | 0.7586 |
| t-stat | 35.25*** | 33.78*** | 33.31*** | 33.87*** | 28.34*** | 30.29*** | |
| HML | Coefficient | 1.0213 | 0.9738 | 1.0941 | 0.9927 | 0.9851 | 1.1131 |
| t-stat | 70.70*** | 59.88*** | 72.83*** | 69.85*** | 52.56*** | 60.60*** | |
| Panel B: Fama–French 5-factor model | |||||||
| Intercept | Coefficient | 0.0150 | 0.0104 | 0.0205 | 0.0128 | 0.0114 | 0.0067 |
| t-stat | 1.42 | 0.86 | 1.89* | 1.22 | 0.79 | 0.48 | |
| MKTRF | Coefficient | 0.9928 | 0.9688 | 1.0271 | 0.9818 | 0.9927 | 1.0089 |
| t-stat | 98.74*** | 83.93*** | 99.28*** | 98.41*** | 70.86*** | 74.68*** | |
| SMB | Coefficient | 0.6146 | 0.6849 | 0.5791 | 0.5856 | 0.6525 | 0.6802 |
| t-stat | 32.26*** | 31.32*** | 29.55*** | 30.98*** | 25.12*** | 27.15*** | |
| HML | Coefficient | 1.2739 | 1.2109 | 1.3780 | 1.2255 | 1.2281 | 1.3898 |
| t-stat | 70.72*** | 58.56*** | 74.35*** | 68.57*** | 50.29*** | 59.01*** | |
| RMW | Coefficient | −0.1105 | −0.0397 | −0.2011 | −0.0857 | −0.0887 | −0.0835 |
| t-stat | −4.39*** | −1.38 | −7.78*** | −3.44*** | −2.57*** | −2.51** | |
| CMA | Coefficient | −0.7344 | −0.7271 | −0.7801 | −0.6862 | −0.7080 | −0.8179 |
| t-stat | −20.82*** | −17.96*** | −21.49*** | −19.61*** | −14.87*** | −17.81*** | |
- Note: This table shows the regression results of Fama–French 3- and 5-factor models for the Forbes Best and Worst Banks sample. The 3-factor model is applied by regressing each portfolio's post-event daily excess returns on market, size and book-to-market factors. The 5-factor model is applied by regressing the excess daily returns for regulated portfolios on a market factor, a size factor, a book-to-market factor, a profitability factor and an investment factor. *, ** and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.
We should interpret the results of the Fama–French models with caution as financial firms are often excluded from asset-pricing tests (e.g., Foerster & Sapp, 2005),10 and additional risk factors may be necessary. For example, Baek and Bilson (2015) examine the issue, arguing that there is no theoretical explanation for excluding financial firms. They empirically confirm this, finding that size and value risk factors are present for financial and non-financial firms. They examine an additional risk factor, changes in interest rates, and find that to be a financial firm-specific risk factor. While the issue does not appear to be fully resolved in the literature, in recent years, some research has employed the Fama–French models without added risk factors, arguing that the models do explain bank returns (e.g., Wansink, 2023).
Next, we examine long-term performance using buy-and-hold abnormal returns (BHARs) (Barber & Lyon, 1997). For each year, we define the holding period as the period from the announcement of the Forbes list until the announcement of the following year's list. The matched sample is the benchmark for BHAR calculations. Therefore, the BHAR is the buy-and-hold return of the Forbes bank sample minus the buy-and-hold return of the matched financial institutions. For each specific subsample, BHARs are aggregated into portfolios containing all securities within a year.
Panel A of Table 5 reports the BHAR results. The Whole sample and all subsamples exhibit negative abnormal returns compared to the matched samples. The BHARs for the Medium Bank and Deteriorate Rank samples are negative and statistically significant compared to their matched samples. Panel A indicates that Forbes-ranked banks exhibit no significantly superior returns compared to matched sample banks. The matched sample exhibits statistically significant superior performance relative to Forbes-listed banks, laying the groundwork for a size, rather than ranking, argument regarding shareholder returns.
| BHR of Forbes sample (1) | BHR of matched sample (2) | BHAR: (1)–(2) | t-stat | |
|---|---|---|---|---|
| Panel A: Matched based on Forbes ranking criteria | ||||
| Best Bank sample | 1.107 | 1.122 | −0.015 | −1.04 |
| Worst Bank sample | 1.147 | 1.149 | −0.002 | −0.11 |
| Medium Bank sample | 1.115 | 1.144 | −0.028 | −1.81* |
| Improve Rank sample | 1.167 | 1.150 | 0.017 | 0.50 |
| Deteriorate Rank sample | 1.158 | 1.234 | −0.076 | −1.99** |
| Whole sample | 1.124 | 1.138 | −0.015 | −1.60 |
| Panel B: Matched based on Forbes ranking criteria and additional criteria | ||||
| Best Bank sample | 1.110 | 1.150 | −0.040 | −2.59** |
| Worst Bank sample | 1.144 | 1.131 | 0.013 | 0.69 |
| Medium Bank sample | 1.108 | 1.132 | −0.025 | −1.60 |
| Improve Rank sample | 1.168 | 1.202 | −0.034 | −0.91 |
| Deteriorate Rank sample | 1.137 | 1.197 | −0.060 | −1.86* |
| Whole sample | 1.121 | 1.138 | −0.017 | −1.80* |
- Note: This table reports the BHARs for the Forbes sample and subsamples. For the calculation of BHAR, we use our matched sample as our benchmark portfolio. Therefore, in this study, BHAR is measured as the return on buy-and-hold investment of our Forbes sample firm less the return on buy-and-hold investment of its matched firm, with differences cumulated over the portfolio. The BHAR holding periods start with the Forbes announcement and continue through the next release date. We evaluate the null hypothesis that the buy-and-hold abnormal returns are equal to zero with the paired t-test. Panel A reports the results using a matched sample based on Forbes ranking criteria. Panel B reports the results using a matched sample based on Forbes ranking criteria and two additional criteria: number of analysts following and institutional holding of the bank. *, ** and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.
Previous literature (e.g., Han & Wild, 1991; Pownall et al., 1993) suggests that analysts are a primary source of investors' information which may affect stocks' long-term returns. In addition, institutional investors may have access to and can analyse the bank's call reports well before retail investors. Therefore, when the Forbes' Best Banks list is announced, this information is already reflected in the stock prices of banks with higher institutional holdings.11 Therefore, to test the long-run buy-and-hold returns, we add two variables to the matching criteria: the number of analysts following the stock and the average institutional holdings of Forbes-listed banks.12 Panel B of Table 5 reports the BHARs compared with the new matching portfolio. The BHARs for the Best Banks, Deteriorate Rank and Whole samples are negative and statistically significant compared to their matched samples.
Filbeck et al. (2013) examine banks listed on the ABA surveys from 1993 to 2009 and find that for the entire survey period and later subperiods based on ROE, the top-performing banks outperform the S&P 500 index on a raw and risk-adjusted basis, including BHARs compared with its matched banks, especially for new-listed banks in the list. In this study, we do not find stock outperformance of the Forbes-listed banks compared with the S&P 500 as our sample consists of the largest 100 banks, and there is little turnover on the list from year to year. Moreover, Filbeck et al.'s matching method is based on bank size, region and loan-to-deposit ratios. We cannot include bank size as a matching criterion due to the nature of the Forbes list. In this sense, the ABA list can be considered more effective in identifying the best-performing banks as it looks at the broader population, not simply the 100 largest banks. In addition, we include the previous year's ROE as a matching criterion (one of Forbes' ranking criteria), which may further confound the results as Gandhi and Lustig (2015) suggest smaller banks outperform larger banks given similar characteristics.
3.3 Long-term accounting performance
The results suggest that no abnormal returns or superior performance are associated with Forbes' press release announcing the Best Banks in the United States. Next, we investigate whether listing as Best or Worst Bank leads to higher accounting performance. We use the matched sample with the two additional matching criteria as the benchmark and compare the change in accounting profitability after being classified as Best or Worst Banks. We use ROA and ROE to measure accounting performance. ROA is the pre-tax income divided by total assets, and ROE is the net income divided by common equity.
The results of the accounting performance of the Forbes' banks compared with their matched samples are reported in Table 6. The results suggest that all but the Deteriorate Rank sample exhibit increased accounting performance. Compared with their matched sample, Best Bank and Deteriorate Bank samples experienced significantly lower changes after being selected by Forbes.
| Sample | Test sample | Matched sample | Test–matched | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | ||||
| Year t−1 | Year t + 1 | (2)–(1) | t-stat | Year t−1 | Year t + 1 | (5)–(4) | t-stat | (3)–(6) | t-stat | |
| Whole sample | ||||||||||
| ROA | 1.14 | 1.30 | 0.15 | 4.70** | 1.03 | 1.26 | 0.23 | 7.46*** | −0.07 | −1.78* |
| ROE | 6.71 | 8.47 | 1.76 | 4.14*** | 5.78 | 7.41 | 1.63 | 1.17 | 0.13 | 0.10 |
| Best Bank sample | ||||||||||
| ROA | 1.51 | 1.54 | 0.03 | 0.61 | 0.94 | 1.25 | 0.31 | 5.76*** | −0.28 | −4.13*** |
| ROE | 9.13 | 9.58 | 0.45 | 1.24 | 3.42 | 8.52 | 5.10 | 2.67*** | −4.65 | −2.42** |
| Worst Bank sample | ||||||||||
| ROA | 0.75 | 1.03 | 0.29 | 4.14*** | 1.13 | 1.29 | 0.15 | 3.00*** | 0.13 | 1.43 |
| ROE | 3.74 | 7.26 | 3.52 | 3.09*** | 7.45 | 8.88 | 1.44 | 2.29** | 2.09 | 1.65 |
| Medium Bank sample | ||||||||||
| ROA | 1.14 | 1.31 | 0.16 | 3.06*** | 1.02 | 1.24 | 0.21 | 3.98*** | −0.05 | −0.79 |
| ROE | 7.14 | 8.51 | 1.38 | 3.15*** | 6.65 | 4.52 | −2.13 | −0.55 | 3.51 | 0.89 |
| Improve Rank sample | ||||||||||
| ROA | 0.96 | 1.38 | 0.42 | 3.67*** | 1.16 | 1.46 | 0.30 | 2.81*** | 0.12 | 0.62 |
| ROE | 5.89 | 9.51 | 3.62 | 3.11*** | 3.57 | 9.83 | 6.26 | 1.31 | −2.64 | −0.57 |
| Deteriorate Rank sample | ||||||||||
| ROA | 1.34 | 1.13 | −0.21 | −1.94* | 0.98 | 1.36 | 0.38 | 3.27*** | −0.60 | −3.93*** |
| ROE | 8.74 | 7.43 | −1.32 | −1.31 | 5.79 | 10.09 | 4.30 | 3.52*** | −5.62 | −3.64*** |
- Note: This table reports the accounting performance for the Forbes sample and subsamples. We use ROA and ROE to measure accounting performance. ROA is the pre-tax income divided by total assets, and ROE is the net income divided by common equity. *, ** and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.
Both the long-term stock and accounting performance results should be interpreted cautiously as being selected is an endogenously determined variable, which could be determined by factors related to firm value. For example, a more profitable bank is more likely to be rewarded with higher stock performance, causing a reverse causality problem. Also, an omitted variable problem may exist.
Next, we investigate whether those banks identified as a Best or Worst Bank lead to higher accounting performance after adjusting for these endogeneity issues. We use the two-stage least squares method to estimate the associations between firm value simultaneously and being selected to the Best and Worst Banks samples.
In Equations (7-10), Rank is defined as 100 minus the Forbes Best and Worst Bank rankings. For example, Rank equals 99 for the highest-ranked bank in the list. Non-performing is the non-performing assets divided by total assets. CAPR1 is the risk-adjusted Tier 1 capital ratio. The efficiency ratio indicates the relationship of a bank's non-interest expense (e.g., salaries, benefits, electric bills and paper clips) to revenue (e.g., net interest income plus non-interest income). Leverage is equal to long-term debt divided by total assets. Growth equals the operating revenue growth rate compared with the previous year. N_Analysts is the number of analysts following the bank. SUE_score is the SUE score, measured by the earnings surprise regarding the number of standard deviations above or below the consensus earnings estimate. Institutional is the institutional holding of the bank. Accounting Performance is measured as ROA and ROE. Improve_Rank is a dummy variable equal to 1 when the bank improves its Forbes ranking by over 20 places compared with the previous year and equals 0 otherwise. Deteriorate_Rank is a dummy variable equal to 1 when the bank worsens its Forbes ranking by over 20 places compared with the previous year and equals 0 otherwise. Log(TA) is the log of total assets, and Rank × Improve (Rank × Deteriorate) is the interaction term of Rank and Improve_Rank (Deteriorate_Rank).
The second-stage regression results are reported in Table 7. We control for fixed-year effects in all models. The results indicate that banks with higher rankings exhibit higher profitability, and banks that deteriorate their rankings from the previous year exhibit lower profitability 1 year after the publication. This finding indicates that being ranked as a Best Bank by Forbes may not suggest better stock performance but rather better accounting performance, as the regression coefficients on Rank in all models are positive. The regression coefficients of Deteriorate Rank in Models (2), (3), (5) and (6) suggest that banks that worsen their Forbes rankings exhibit lower profitability 1 year after the publication. This result implies that if an investor monitors the listed banks, their rankings and the movements up or down in ranks year-over-year, they can identify banks with better (or worse) accounting performance in the coming year.
| Model (1) | Model (2) | Model (3) | Model (4) | Model (5) | Model (6) | |
|---|---|---|---|---|---|---|
| ROE t+1 | ROE t+1 | ROE t+1 | ROA t+1 | ROA t+1 | ROA t+1 | |
| Intercept | 9.6053 | 9.6394 | 9.7027 | 1.7539 | 1.7701 | 1.7809 |
| (3.85***) | (3.87***) | (3.89***) | (6.54***) | (6.60***) | (6.63***) | |
| RANK t−1 | 0.0338 | 0.0308 | 0.030 | 0.0058 | 0.0056 | 0.0054 |
| (5.89***) | (5.15***) | (4.82***) | (9.07***) | (8.31***) | (7.81***) | |
| Improve_RANK t−1 | 0.1438 | 1.3003 | −0.0506 | −0.0478 | ||
| (0.28) | (0.80) | (−0.91) | (−0.28) | |||
| Deteriorate_RANK t−1 | −1.032 | −2.1807 | −0.1093 | −0.2037 | ||
| (−1.84*) | (−2.23**) | (−1.83*) | (−1.95*) | |||
| RANK × Improve | −0.0195 | 0 | ||||
| (−0.73) | (−0.01) | |||||
| RANK × Deteriorate | 0.0404 | 0.0033 | ||||
| (1.45) | (1.11) | |||||
| Log(TAt) | −0.0448 | −0.0372 | −0.0441 | −0.0586 | −0.0591 | −0.0599 |
| (−0.21) | (−0.17) | (−0.20) | (−2.55**) | (−2.56**) | (−2.60***) | |
| ROE t−1 | 0.0635 | 0.0726 | 0.0727 | |||
| (2.07**) | (2.33**) | (2.33**) | ||||
| ROA t−1 | 0.107 | 0.1103 | 0.1105 | |||
| (3.58***) | (3.59***) | (3.58***) | ||||
| N_Analysts t−1 | 0.1225 | 0.1163 | 0.1176 | 0.019 | 0.0187 | 0.0189 |
| (3.29***) | (3.11***) | (3.14***) | (4.79***) | (4.68***) | (4.72***) | |
| SUE_Score t−1 | 0.0618 | 0.0619 | 0.0609 | 0.0087 | 0.0088 | 0.0087 |
| (2.22**) | (2.23**) | (2.19**) | (2.94***) | (2.97***) | (2.94***) | |
| Institutional t−1 | −0.0098 | −0.0077 | −0.0077 | −0.0032 | −0.0029 | −0.0029 |
| (−0.98) | (−0.77) | (−0.77) | (−2.96***) | (−2.69***) | (−2.69***) | |
| Year fixed effect | YES | YES | YES | YES | YES | YES |
| R-squared | 0.183 | 0.186 | 0.189 | 0.268 | 0.271 | 0.272 |
- Note: This table shows the second-stage regression results for firm performance ranked by Forbes Best Banks. We use two-stage instrumental variable regressions to control for endogeneity. Our regression models are reported in Equations (8) and (9). We use three measures of firm performance: return on assets (ROA), return on equity (ROE), and annual raw return. Improve_Rank (Deteriorate_Rank) is a dummy variable equal to 1 when the bank improves (worsens) its Forbes ranking by over 20 places compared with the previous year and equals 0 otherwise. Log(TA) is the natural log of the bank's total assets. N_Analysts is the number of analysts following the bank. SUE_Score is the SUE score, measured by the earnings surprise regarding the number of standard deviations above or below the consensus earnings estimate. Institutional is the institutional holding of the bank. All models are controlled for year-fixed effects. *, ** and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.
We conduct a placebo test to check the validity of the results in Table 7. Specifically, we randomly assign a ranking between 0 and 99 for each year and re-estimate the principal regression, as shown in Table 7. We repeat the process 500 times. The unreported test results show that t-statistics for Rank are insignificant from the 500 estimates for all six models. This result suggests that confounding effects of omitted variables or endogeneity issues do not cause the results reported in Table 7.
There is limited evidence that banks exhibit superior short-term or long-term stock performance. There is some evidence that accounting performance may change after a movement (up or down) on the Forbes' Best Banks list. One possible explanation is the size effect of banks. Gandhi and Lustig (2015) examine the relationship between bank size and risk-adjusted returns and find that smaller banks' risk-adjusted returns are higher than those of larger banks. Since the Forbes survey considers only the largest 100 banks in asset size, and the matched samples are consistently smaller compared with the Forbes' bank samples, it is not surprising that we do not find significantly higher returns in stock performance, especially when compared with the matched samples (e.g., BHAR results). In addition, since these largest banks are usually associated with a greater number of analysts following the banks and higher institutional holdings, the inaccessibility of call reports to the retail traders may not cause an issue for these largest 100 banks. Therefore, the additional information in the Forbes' listing may have already been reflected in these banks' stock prices.
That being said, the Forbes listing contains some additional value in predicting the accounting performance of listed banks, especially when compared with the other banks in the list and their past performance. One possible explanation is the efficiency of the stock when dissecting information. The market quickly reflects information in bank stock performance, while accounting performance may lag.
4 CONCLUSIONS
This paper investigates the impact of the Forbes's Best Banks survey on the share price response of listed banks and BHCs and the longer-term investment performance of listed institutions. Mixed results are associated with the initial announcement effect as smaller banking institutions benefit most from the surveys' release. We find positive and statistically significant announcement effects in the window surrounding the announcement, but the impact is negated in the following days. Except for the Whole sample, all subsamples exhibit statistically insignificant negative returns over the entire event window.
The longer-term financial implications of the survey are limited. The overall sample and each subcategory show statistically insignificant performance differences compared with the S&P 500 index. Risk-adjusted performance measures are mixed and lack statistical significance. In the short run, it appears that the market does not see value in gaining access to publicly available data, even though it is relatively inaccessible. After controlling for fixed-year effects, banks with higher Forbes rankings exhibit lower annual raw returns. However, banks with higher rankings or those that see improved (deteriorated) rankings from the previous year exhibit improved (reduced) profitability in the following year. This finding has interesting economic implications. In finance, the adage is that the past doesn't predict the future. However, in this case, the bigger the move up/down the ranking, the better/worse the bank performs in the following year. This discovery suggests that Forbes is ‘seeing the future’ regarding accounting performance.
From an investor's perspective, the economic significance of the results rests primarily on investors monitoring the year-on-year changes in the survey rankings. Investors seeking a long-term strategy to profit from the Forbes Best Banks survey should recognise that firms that move up (down) in the rankings exhibit improved (decreased) profitability in the subsequent year. Otherwise, the market seems ambivalent about this information based on the lack of consistent short-term announcement performance or longer-term holding period performance.
While examining the benefit to investors using the Forbes analysis and subsequent rankings to make investment decisions was the first goal of this study, another finding emerged. Using a different and even unintended lens, we were able to reinforce the size effect in banking. Gandhi and Lustig (2015) examine the relationship between bank size and risk-adjusted returns and find a significant relationship between the two – smaller banks outperform larger banks. The Fama–French models argue that size is critical to understanding asset pricing. We have inadvertently looked at size using our matching methodology. First, the Forbes survey considers only the largest 100 banks in asset size. We then consider returns to shareholders based on the Forbes Best (smaller banks on the Forbes list) and Worst banks (usually larger banks on the Forbes list). We also match to a sample of banks not included in the survey. As we didn't include a bank that had ever been listed, we, by definition, matched to smaller asset-sized banks. To our initial surprise, the matched sample banks outperformed the Forbes-listed banks. However, it became clearer upon examining the descriptive statistics of the various samples that the matched sample banks are about half the size of the whole Forbes sample (i.e., approximately $59 billion versus $129 billion). While the approach differs from Gandhi and Lustig's examination of size and bank performance, we arrive at the same conclusion, supporting the assertion that smaller banks outperform larger banks on both an accounting and long-term return basis. Compared with the results from ABA bank surveys (e.g., Filbeck et al., 2013, 2016), we find insignificant results in the Forbes-listed banks' stock performance, especially when compared with the matched portfolios, and we believe that one of the important determining factors is the bank size. We conclude the ABA list is more effective in identifying the best-performing banks as it looks at the broader population. Based on our results, we suggest that Forbes consider reporting/ranking the best and the worst performing banks for medium and small banks in their future surveys to have a more comprehensive picture of the impact of the ranking criteria of surveys on bank performance. We believe this provides an alternative mechanism for examining bank size and performance and warrants further research.
Open Research
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
REFERENCES
- 1 Forbes continues to publish the survey each year.
- 2 The Compustat Bank Annual file contains annual accounting data on approximately 638 of the leading United States banking institutions.
- 3 We also evaluate subsamples with 5 or 10 ranking changes compared with the previous year. The results are qualitatively similar.
- 4 Unreported t-test results show that the differences are statistically significant at at least the 5% level.
- 5 Unreported t-test results show that the difference is statistically significant at the 1% level.
- 6 Unreported t-test results show that the differences are statistically significant at at least the 5% level.
- 7 Unreported t-test results show that the difference is statistically significant at the 1% level.
- 8 Event study results for event window (−5, +5) are qualitatively similar and therefore omitted for brevity.
- 9 We evaluate the event study results using different ranking categories (e.g., non-performing assets/total assets, efficiency ratio, operating revenue growth) and find similar results.
- 10 We thank an anonymous referee for this comment.
- 11 We thank an anonymous referee for suggesting these two additional control factors.
- 12 Note that both the number of analysts following the stock and institutional holdings are positively correlated with bank size (test results are unreported for brevity). We do not use this matching criterion in our original matching portfolio as these two additional matching criteria significantly reduce the sample size when using the propensity score matching method.
