Please Note: Blog posts are not selected, edited or screened by Seeking Alpha editors.

A Note On The Ability Of Banks Sustain Dividends (Jan 22, 2012)

A Note On The Ability Of Banks Sustain Dividends (Jan 22, 2012)

Recent financial crisis (ongoing from 2007 in the author mind) gives us a nice opportunity to study sustainability of dividends of different firms. I consider banks only in this paper.

There are hundreds banks in the USA traded at stock markets. Most of them are small banks that serve local communities. Usually capital appreciation of such banks' stocks is quite modest and investors rely on dividends. Many US banks paid steady or even growing dividends (SGD) before The crisis divided the banks into 2 large groups - one consists of banks that continue to pay SGD (no cuts) and another consists of banks that suspended or reduced (sometime very strongly down to 1 cent) dividends (labeled as cutters). The research goal is to figure out if cutter and non-cutter banks have a difference at the time just before dividend reduction from public financial reports.

It seems important to figure out the possibility and probability of a dividend cut for a particular bank. In order to perform this task different metrics have been evaluated by the author for banks from these 2 groups. The metrics are not suppose to be very precise but should be able to give a rough picture of the situation.

Each publicly traded US bank reports financial data reported on regular basis. In these reports (as well as in more detail annual reports) bank management rounds numbers (e.g. loans amount) usually by thousand dollars or so. Hence a history of any reported number can be considered as a discrete dataset. Although the author doesn't think that quantum mechanics is a suitable theory for financial world, it seems possible to imagine that time changes of reported data can be presented as quantum fluctuations and to use appropriate mathematics. One of the proprietary metrics named B4S is constructed in analogy with quantum fluctuations. [Just if you are curious - the letter B in B4S metric name reflects that it is good for banks (it does not work for industrial companies but was not tried for other financial firms), 4 because it is the fourth author's metric for banks and S means simplified (it can be improved for more detail research)]. B4S values have been calculated for each bank that pay SGD or paid SGD before the cut. Amount of past dividends determine partially B4S value along with other variables commonly cited in quarterly and annual financial reports.

B4S suppose to show the difference between banks that continue to pay SGD and banks that reduced dividends in a few quarters before and after the dividend cut as it outlined in figure 1. B4S value should increase significantly after the dividend cut event because of strong reduction (and even elimination) of dividends during recent financial crisis. It is expected that B4S value should decrease before the dividend cut probably as early as 3 years or as late as 1 quarter before the cut marked by zero in the X-axis of figure 1 for group of banks that reduce dividends (cutters). These time periods expectations based on results of prior scholar studies for dividend cuts for industrial companies. It is expected that B4S value should not significantly change for group of banks that do not reduce dividends. For both groups, mean and median values are expected to be different (distributions of individual banks B4S skewed to high values) and fluctuations around median are expected to be are relatively strong.

The data from 1 January 1996 till 1 January 2012 have been analyzed for about 300 banks. The author mostly rely on data for dividends but it seems that pre-2007 dividend histories were erased from their database for some (seems mostly financial) companies. [Another better free source of or dividend histories is but their data are often redundant and author did not figure out how make computer to handle such situation (help is appreciated)]. Although quarterly financial reports are available at SEC web site, secondary sources of information were used. The author did not check reliability of financial information from these sources.

Banks with semiannual and annual dividends and banks with strongly controversial financial data from different Internet sources were excluded. The banks that reduced regular dividends within 1 year after a special dividend (that was at least 10% higher that their regular dividend) were not considered as cutters. Only the first cut of dividends occurred between 1996 and 2012 was considered. Several banks further reduced or omitted dividends after the first cut, but these events were not considered.

B4S metrics have been calculated for 140 banks that reduced their dividend. The times of dividend cuts were approximated by ex-dividend date collected from and by comparison of dividends amount reported in consequence quarters by banks. More precise data should operate with dividend announcement dates. B4S metrics have been calculated as well for 70 banks that did not reduced dividends during the recent financial crisis. Most of these banks are taken from David Fish's CCC list of companies that increased their dividends in last 5 years or longer ( The banks with at least 5 years of steady (constant) dividends are also considered. In cases when dividend data are missed in the banks quarterly reports, conformations from annual reports and data have been obtained.

In order to unify the picture the time series for financial data of individual cutters were shifted to the quarter when most of banks reduced dividends (second quarter of 2009), hence the reported quarter 0 in the X-axis of figure 2 corresponds to June 2009. The shift does not affect results and just help to improve visualization. B4S metrics have been calculated for 9 quarters after the dividend reduction. If the cut occurs quite recently all financial numbers for "missing" quarters considered equal to zero.

The median B4S values shown for these two groups of banks in figure 2 are quite surprising although some expectations are confirmed.

The expected results are considered first. Median value of B4S for the group of banks that reduced dividends indeed increased rapidly after the cut but then reduced because not all banks have enough data. The increase started at the quarter -1 because ex-dividend dates are used instead of dividend announcement dates and amount of dividends decreases in bank quarterly reports after the pay dates. These time delays embedded into the figure 2 lead to the spread of points within 1 quarter around the dividend cut date.

Median value of B4S for the group of banks that continue to pay SGD during the recent financial crisis did not change significantly around the quarter 0. Fluctuations of B4S medians with time are strongly for not cutters because these group in 2 times smaller than the cutters group. Fluctuations of B4S medians are higher for quarters -50 to -23 before the cut that for quarters -22 to +9 before and after the cut because older financial data are missing for some banks.

B4S medians for cutters were below 2.1 for 8 quarters before the dividend reduction (see figure 3 which is a part of the figure 2). B4S medians for cutters dropped below 2.0 for 7 quarters before the dividend reduction if the quarter -1 is excluded. During the remaining 42 quarters of data the B4S medians for cutters dropped below 2.1 only for two other quarters (i.e. below 5% of events). The average of B4S medians for the cutters group before the dividend cut was about 2.5 and the maximum of B4S medians for these banks was below 3.22.

Absolute minimum of B4S medians for no-cut banks was slightly above 2.3 as shown in the figure 3 (which is a part of figure 2). In all other quarters B4S medians for no-cut banks only twice were below 2.5 (again only about 5% of events), while the average of B4S medians for this group was about 3.5.

The most unexpected result is following. The difference between the 2 groups of banks that did and did not reduce dividends is visible as early as -33 quarters before the dividend cut (see figure 2). During these 8 years an investor who is concern about dividend income had a plenty of opportunities to sell cutters stocks at relatively high prices and to buy no-cut banks at relatively low prices and to collect stable dividends regardless the financial crisis. (Unfortunately, the author was not such smart investor and actually lost income and money on some banks shares he owned.)

B4S seems not absolutely perfect metric because some banks such as CenterState Banks, Inc. (NASDAQ:CSFL) or TriCo Bancshares (NASDAQ:TCBK) [other are NRIM, SIFI, UCBI) ] reduced or omitted dividends even having minimum B4S above the maximum of B4S medians for banks that decreased dividends (equal 3.22). It seems interesting to figure out if dividend cuts of these banks were caused by their financial situations or were caused by human factors (e.g. these banks management just followed the crowd or regulators forced these banks reduce dividends regardless their ability to sustain pre-crisis dividends amounts).

Other banks such as New York Community Bancorp, Inc (NYB), First Niagara Financial Group (NASDAQ:FNFG) and Camden National Corporation (NASDAQ:CAC) have very low average B4S but keep steady dividends during the financial crises. See Notes 1 and 2.

Therefore type 1 and type 2 errors for B4S are small. See Note3.

Of course the value of B4S metric for future investments is hard to estimate now because it can be

a) result of specific conditions of the financial crisis. But the selection of the financial parameters for the metric construction was rather related with common sense about bank operations (as well as the author with very limited economics and finance knowledge understand banks) and not related to the crisis. Limited number of dividend cuts for few banks prior year 2007 had the same pattern for B4S metrics as banks that reduced dividends during the recent financial crisis.

b) result of data mining. But the physical and mathematical model built for B4S did not use any data mining technique (the author simply doesn't really know these techniques) and no prior history was used for the selection of the financial parameters for B4S metrics.

B4S metric itself does not give answer on probably main question - what happens with banks about 8 years before they have to cut dividends? One unproven hypothesis is to consider mismanagement of such banks, for example possible switch from traditional risks estimations to popular at that time copula function approach or difficulties to handle more competitive environment after repeal of the Glass-Steagall Act (Gramm-Leach-Bliley Act) at the end of 1999.

In order to check robustness of the results cutters were split into two equal sets (70 banks in each set). There are different possibilities for the split - a simple alphabetical one was used. Median B4S are shown in the figures 4 and 5, where the figure 5 is again a fragment of the figure 4.

The obtained result confirms main conclusions of the overall tests.

It seems interesting to figure out how B4S metric works for other financial firms. The results will be reported elsewhere.

Disclaimer: At the moment the author own stocks of several US and foreign banks but not the stocks spelled out in the article. The author doesn't know that market will do in next 72 hours and cannot promise not to act if something good or bad happens.

Additional disclosures: I submitted this as SA article 22 January 2012, it was rejected.

1) FNFG reduced dividends 25 Jan 2012.

2) The author neglected some financial parameters for B4S construction which might cause abnormal B4S values for these banks.

3) Type 1 error is classifying bank that cuts dividends as a non-cutter, and type 2 error is classifying bank that did not cut dividends (i.e., non-cutter bank) as a cutter.