Long-run performance of Initial Public Offerings of banks

(1)

Long-run performance of Initial Public

Offerings of banks

Studentnr: s2168146 Name: Roeland Wilmink Study Program: MSc Finance Supervisor: Prof. Dr. Wolfgang Bessler Field Key Words: IPO, Banking & Insurance

Word count: 14,509

Abstract

(2)

2

I. Introduction

The performance of initial public offerings has been widely documented and researched for over decades. There are several reasons why the long-run performance of IPOs in the banking sector is of interest. First, the outcome of this research can show the presence of price patterns which can be used by investors for their trading strategies. This can result in higher returns for their investments. Second, the outcome can indicate the efficiency and accuracy of the IPO market. Underperforming the long run and vice versa will indicate that this market is not working as it should which can be taken into consideration when investing. Third, the time window of the research with the combination of the banking sector is interesting since the recent developments during the financial crisis, including governmental takeovers, can have an impact on the IPO performance. Hence, this research adds to existing literature by examining a more recent window of IPO performance in a detailed way.

This paper uses previous research of Ritter (1991) which examines the performance of IPOs for several industries. He finds that IPOs underperform the market in general while IPOs of financial institutions outperform the market and he concludes that banking IPOs differ from non-banking IPOs. This research focuses on banking IPOs and finds that they outperform the market over a period of 3-year. However, size and growth effects are not found in the used dataset even though other literature predicts such effects. European IPOs underperform slightly in one case.

First, the existing literature on this topic will be discussed followed by a description of the data and the used methodology. Next, the results will be presented along with robustness checks of the used method. Finally, a discussion and conclusion about the research will be presented.

II. Literature

(3)

3

initial public offerings regarding 1,526 IPOs that went public in the U.S. during the 1975-1984 period. He found that this sample of IPO stocks had a return of 34.47% in the 3 years after going public using a holding value calculation (BHAR) from the closing market price of the first day of trading and that of its 3 year anniversary. This result is compared to that of a control sample with 1,526 listed stocks matched by industry and market value. This control sample produced an average return of 61.68% during the same 3 year holding period which translates into a ratio of approximately 0.831 for every dollar invested in the IPOs in comparison to the control sample. So, over a period of 3 years, the IPOs underperformed. In his paper, Ritter (1991) offers several possible explanations for this underperformance: risk mismeasurement, bad luck or fads and over optimism. Various cross-sectional and time-series patterns are presented in order to establish which explanation is most likely for this result of underperformance. One of his findings is that underperformance occurs more among young growth companies. Also, he found that companies that go public in high-volume years in terms of IPOs tend to have performed worse in the 3-year window. This would indicate that momentum and investors being too optimistic about the future performance can have an impact on the IPO return. This finding is supported by Lee, Schleifer and Thaler (1991) who find that the annual number of companies going public is negatively related to a self-made measure of individual investor sentiment, based on the discount on closed-end mutual funds for a sample during the period 1966-1985. Thus, issuers time their offers in order to lower the cost of capital for the company.

(4)

4

This makes it a topic of interest to research what has happened to similar IPOs of financial institutions under different circumstances.

For banks, a special approach is proposed by different papers. Bessler, Murtagh and Siregar (2003) find evidence that the dividend policy of bank IPOs differ from that of non-banks. Non-financial firms or industrial firms do not start paying dividends immediately after going public because they have a high cash flow demand for the investment opportunities that these firms face. On the contrary, banks do start paying dividends early on because the market faces information asymmetry for this particular sector which calls for banks to reveal their true financial condition. Banks do so by signaling in forms of paying dividends to their shareholders. For non-banks, paying dividends signals that the firm has good investment projects and is able to generate cash flows. In addition, over-investing is countered since higher dividends decrease the available funds to the over-investing management. Paying out dividends has a positive impact on the valuation of the firm by the market. For banks, a change in financial policy such as dividend is a signal of the quality of the portfolio loan to investors (Bessler and Nohel, 1996 and 2000). An increase in dividends results in positive stock price reactions while a dividend decrease results in negative stock price reactions for both the short and long run. The differences in dividend policies provides an additional argument to distinguish the IPOs of banks and non-banks.

(5)

5

have a better governance system which results in the reasonable decision to find a bidder as an exit strategy.

Besides this, Francis, Hasan and Siregar (2009) argue that some bank characteristics can indicate whether the bank goes public (IPO) or is taken over (M&A). They find evidence that fundamental firm factors such as age, size and liquidity ratio along with market timing are important variables that influence this decision or outcome. Size is found to be an important explanatory variable by Bessler et al. (2003) as well as larger banks underperform the non-IPO bank index by -20% while smaller banks match this benchmark for the five-year holding measure. Moreover, a reduction in geographic restrictions such as allowing for interstate banking makes it more likely that banks will perform or are subject to a takeover. This research also suggests that self-selection exists when banks choose for an IPO or M&A strategy which can cause a bias.

(6)

6

Interestingly, Ibbotson (1975) calculates the long-run performance for companies that went public for the period 1960-1969 using a performance window of 5 years. His research does not find statistically significant deviations that confirm market inefficiency in the aftermarket which is found by Ritter (1991) for a period of 3 years. However, Ibbotson does find that there is a positive performance in the first year and the fifth years while there is a negative performance in the years 2, 3 and 4. This suggests that a different definition of long-term performance can deliver different outcomes. Consequently, Loughran and Ritter (1995) take a 5-year window for IPO long-run performance based on the finding of Loughran (1993) who reports that IPOs underperform. They find that an IPO is a poor investment for the 5-year period with an annual return of 5% compared to a market portfolio with an average annual return of 12%. So, it is possible that the performance of an IPO depends on the chosen time window in terms of timing.

In line with this, Loughran and Ritter (1995) propose that investors have hopes that are too high as an explanation of the ongoing misjudgment of the returns of IPOs. They suggest that a longer period of time is necessary for investors to conclude that their probabilities of finding a superb IPO have been too high in the past. This can be analyzed by looking at a more recent time window than earlier studies during the nineties. Then, the research question that is being evaluated by this paper is: How do the IPOs of banks perform in the long-run?

Hypothesis 1: The IPOs of banks outperform the market in the long run

Hypothesis 2: Larger size and higher growth of a bank will affect the performance in a negative way

III. Data & Methodology 1. Data

(7)

7

with an average share value that is lower than $1 during the examined period of 36 months are removed which results in 115 observations that can be used to assess the long-run performance. The observations and its data are examined, banks with a high degree of incomplete or inconsistent data are removed before winsorizing is used at a 1% for the stock returns to remove possible outliers. A total of 86 observations from 23 countries (Table I, appendix) is suitable to be studied regarding the long run performance. Out of the 86 banks, 2 banks were merged during the 3-year window while an additional 10 were merged after the window ended. One bank was liquidated after the 3-year window. Only the two banks that were merged during the 3-year window might affect the outcome which results in a negligible survivorship bias effect. The returns and data until the last month of public activity are taken into account, after that there is no data that can be used for calculation which results in a drop of used banks in the sample from 86 to 84. Table II shows the distribution of the IPOs per year and the average size in terms of total assets. There is a decrease of amount of IPOs after 2007 which might be caused by the financial crisis that started during 2007. Out of the 3,182 stock price observations, 60 monthly prices are missing of which 26 are due to a merger after which no more stock price data is available and 34 observations are caused by incomplete data. The latter group exists of missing stock price data (32 observations) and missing IPO prices (2 observations). The 32 observations caused by missing stock price data are recalculated using the linear growth method between two known observations.

Table II: Yearly distributions of IPOs and size

Year Frequency Size (assets in mln$) _{Size (total loans in mln$)}

Nr. Obs. % of total Mean Median _Mean Median

2007 41 47.67% 17,476 2,791 _11,553 _1,798 2008 8 9.30% 16,222 4,037 _14,819 _3,102 2009 10 11.63% 10,892 9,064 _5,830 _4,477 2010 8 9.30% 5,168 2,060 _3,056 _1,278 2011 10 11.63% 14,906 4,467 _10,152 _2,621 2012 9 10.47% 9,839 5,411 _3,755 _1,838

Total 86 100% Avg. p.I. 14,351 4,038 Avg. p.I. _9,421 _2,282

The left table presents the number of observations in terms of IPOs per year. The middle table shows the mean and median of the total assets in million US dollars per year of the banks that went public. The right table shows the mean and the median of the total loans in million US dollars per years of the banks that went public. The lower rows show the total observations, average mean and median per IPO for both total assets and total loans.

(8)

8

banks are put in a portfolio to test whether there are abnormal returns in comparison to the overall market, measured with the MSCI excluding the US and Canada index. Additionally, the returns are calculated using a tracker, iShares Global Financial, which measures the world wide performance of several financial institutions. This combination of indices will provide an insight of the IPO performance both in comparison to the general and financial specific market (Graph I, appendix). Next, the sample is split up in several areas around the world. The US, Europe, Latin-America and Asia are used as an area and the country origin of the IPO in terms of listing is used and combined in order to construct portfolios (Table III, appendix). These areas all have their own market index as benchmark which also takes into account their size as the sample is split in a subsample with large (37 observations) and small banks (38 observations). In addition, both the large and small samples are tested to see whether there is a possible size effect. The banks are ranked regarding their size which is based on a combination of both total loans and total assets.

Table IV: Descriptive statistics of the abnormal returns per used benchmark (market index)

The amount of monthly observations, the mean, the median, standard deviation, minimum, maximum, skewness and excess kurtosis of every used index are presented column wise. The MSCI ex US & Canada index is used as the general market index, the iShares Global Financial is used as the general, financial market index. The S&P banks is used as the index for large US banks, S&P regional for small US banks, Stoxxx Eur is the index used for large European banks, Lyxor MSCI EMU SC is used for small European banks. MSCI Brazil Capped is used for Latin American large banks, MSCI Pacific ex Japan is used for large Asian banks and SPDR International Small Cap is used for both small Latin American banks and small Asian banks.

(9)

9

stable than the financial market index. The mean of the MSCI index is closer to one as is its median, its standard deviation is lower which implies lower volatility and in addition, the minimum and maximum values are less extreme.

The descriptive statistics of the average abnormal returns are presented in Table V. The financial market abnormal returns have the highest median, the Latin American abnormal returns the highest mean while the Asian abnormal returns have the lowest mean, median and minimum. The Latin American abnormal returns are the most volatile given the standard deviation of 5.13% and have the highest maximum. The distribution should also be taken into account since the skewness of 0.9213 and excess kurtosis of 2.0164 imply that it might be non-normal for the financial market abnormal returns.

Table V: The amount of monthly observations, the mean, the median, standard deviation, minimum, maximum, skewness and kurtosis of the average abnormal return are presented column wise as. The MSCI ex US & Canada index is used as the general market index, the iShares Global Financial is used as the general, financial market index. The S&P banks is used as the index for large US banks, S&P regional for small US banks, Stoxxx Eur is the index used for large European banks, Lyxor MSCI EMU SC is used for small European banks. MSCI Brazil Capped is used for Latin American large banks, MSCI Pacific ex Japan is used for large Asian banks and SPDR International Small Cap is used for both small Latin American banks and small Asian banks.

Index Nr. Obs. Mean % Median % Std. Dev % Min % Max % Skewness Kurtosis General market 36 0.0058 0.0053 0.0188 -0.0416 0.0464 0.1106 0.5183 Financial market 36 0.0086 0.0078 0.0156 -0.0201 0.0564 0.9213 2.0164 Large 36 _0.0046 _0.0008 _0.0224 _-0.0334 _0.0555 _0.3713 _-0.5074 Small 36 _0.0026 _0.0010 _0.0217 _-0.0420 _0.0531 _0.3581 _0.0072 US 36 _0.0055 _0.0039 _0.0189 _-0.0338 _0.0406 _-0.1498 _-0.5341 EU 36 _0.0004 _0.0003 _0.0265 _-0.0451 _0.0639 _0.2719 _-0.4232 Asia 36 _-0.0050 _-0.0046 _0.0513 _-0.0886 _0.1259 _0.5794 _0.0731 Latam 36 _0.0105 _0.0070 _0.0520 _-0.0684 _0.1763 _0.9368 _1.4801 2. Methodology

Ritter (1991, 1995) offers two measures that can be used to evaluate the long-run performance. The first one being the average abnormal returns (AR)1 and the cumulative average abnormal returns (CAR). Second, a 3-year buy-and-hold strategy is used to calculate the return of the IPOs. There are several pros and cons regarding the use of these measures therefore, this paper assesses both (Gur-Gershgoren, Hughson, Zender, 2008). Monthly

1_*

(10)

10

benchmark-abnormal returns are calculated as the monthly return minus the monthly benchmark return for the corresponding return. This benchmark- abnormal return for stock i in month t is calculated as:

𝑎𝑟(𝑖, 𝑡) = 𝑟(𝑖, 𝑡) − 𝑟 (𝑚, 𝑡) (1)

The average benchmark- abnormal return on a portfolio of n stocks for month t is the weighted average of these individual benchmark- abnormal returns:

𝐴𝑅𝑡 = (_𝑛1) ∗ ∑𝑛_𝑖=1𝑎𝑟(𝑖, 𝑡) (2)

The cumulative benchmark- abnormal returns aftermarket performance, used to assess the significance of the returns, from event month q to event month s is the summation of the average benchmark- abnormal returns and can be defined as:

𝐶𝐴𝑅(𝑞, 𝑠) = ∑𝑠 𝐴𝑅𝑡

𝑡=𝑞 (3)

The t-statistics for both the abnormal returns and cumulative abnormal returns are calculated as executed by Ritter (1991). The t-statistic for the abnormal returns is computed as:

𝑡 𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 = 𝐴𝑅𝑡 × √𝑛/𝑠𝑑𝑡 (4)

Where ARt is the average abnormal return of the firms at month t, n is the amount of observed firms per month and sdt is the cross-sectional standard deviation of the returns for month t. The t-statistic for the cumulative abnormal return in month t (CARt) is calculated as:

𝑡 𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 = 𝐶𝐴𝑅𝑡 × √𝑛/𝑐𝑠𝑑𝑡 (5) N is the amount of observed firms per month t, csdt is computed as:

𝑐𝑠𝑑𝑡 = [𝑡 × 𝑣𝑎𝑟 + 2 × (𝑡 − 1) × 𝑐𝑜𝑣]1/2₍₆₎

Where t is the event month, var (over 36 months) is the average cross-sectional variance and cov is the first-order autocovariance of the ARt series.

An alternative method to evaluate the buy and hold strategy return can be defined, the 3-year holding returns (BHAR) can be used:

𝑅𝑖 = ∏36 (1 + 𝑟(𝑖, 𝑡))

(11)

11

Here, r(i,t) is the raw return of firm i in month t which implies that the total return of a stock purchase at the first closing market price after going public and held until the 3-year anniversary of the IPO or its delisting is measured. The BHAR measure allows for compounding whereas the CAR does not as the BHAR employs geometric returns rather than arithmetic returns. The BHAR is calculated per month per company (Ri) and the market index as expected return is subtracted in order to get the BHAR per month:

𝐵𝐻𝐴𝑅(𝑖) = [∏_𝑡=𝑇1𝑇2(1 + 𝑅𝑖, 𝑡) − 1] − [∏_𝑡=𝑇1𝑇2(1 + 𝐸(𝑅𝑖, 𝑡)) − 1] (8)

Where E(Ri,t) is the expected return as presented with a selected market benchmark. Next, the BHARs of the 86 banks are taken after 36 months of calculation and weighted and summed to get an average BHAR:

𝐵𝐻̂𝐴𝑅 = (_𝑛1) ∗ ∑𝑛_𝑖=1𝐵𝐻𝐴𝑅(𝑖) (9) The t-statistic for the buy and hold returns is computed as:

𝑡 𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 = 𝐵𝐻̂𝐴𝑅/ (𝑠𝑑

√𝑛) (10)

Sd is the cross-sectional standard deviation of the total buy and hold returns after month 36, n is the amount of observed firms.

In addition, a multiple regression model is estimated to determine whether the proposed variables such as size and implications of the AR and CAR test outcomes are similar. The model assesses 44 firms (Table VI, appendix) from the total sample of which both sufficient stock price data and accounting is available. Large firms, in terms of assets, underperform smaller firms (Bessler et al., 2003) while high growth banks also underperform low growth banks which can be seen by increasing loan loss provisions for the high growth banks (Houge and Lougran, 1999). The loan loss provisions as a percentage of total loans are plotted in Graph II (appendix) where an increase is seen after year 2 and 3. The variables will be tested using an OLS model in a similar way as Ritter (1991) to determine if they affect the returns. The model will assess whether assets, loans, growth, loan loss provision, area and timing of the IPO have an effect on the abnormal average returns. The model is specified as:

𝑅𝑒𝑡𝑢𝑟𝑛 𝑖 = 𝑏0 + 𝑏1 𝐿𝑜𝑔 (𝑎𝑠𝑠𝑒𝑡𝑠 𝑖) + 𝑏2 𝐿𝑜𝑔 (𝑙𝑜𝑎𝑛𝑠 𝑖) + 𝑏3 𝑎𝑠𝑠𝑒𝑡 𝑔𝑟𝑜𝑤𝑡ℎ 𝑖 + 𝑏4 𝑙𝑜𝑎𝑛 𝑔𝑟𝑜𝑤𝑡ℎ 𝑖 + 𝑏5 𝑝𝑟𝑜𝑣𝑖𝑠𝑖𝑜𝑛 𝑓𝑜𝑟 𝑙𝑜𝑎𝑛 𝑙𝑜𝑠𝑠𝑒𝑠 𝑖 + 𝑏6 𝐸𝑈𝑖 + 𝑏7 𝑈𝑆 𝑖 +

(12)

12

Return i is the sum of the monthly average abnormal returns per bank and per year. This will result in average returns for 3 years after the IPO. These 3 values are averaged to get the average return per bank for the year period. Log (assets i) is the natural logarithm of the 3-year average total assets stated on the balance sheet after the IPO per individual bank. Log (loans i) is the natural logarithm of the 3-year average total loans stated on the balance sheet after the IPO per individual bank. Asset growth i is the average asset growth from t -1 (1 year before the IPO) until the end of the 3-year period per individual bank based on total assets reported on the balance sheet. Loan growth i is the average loan growth from t-1 until the end of the 3-year period per individual bank based on total loans as reported on the balance sheet. Provision for loan losses i is the amount of total loan loss provision as reported divided by the total amount of loans as reported in the financial statements. EUi, USi, and Asiai, are dummies for the areas in which the IPO took place with value 0 if the IPO was not in that area and value 1 if it was in that area. 2007i, 2008i, 2009i, 2010i and 2011i are time related dummies with value 0 if the IPO did not take place in that year and value 1 if the IPO took place during that year.

IV. Results

1. Stock performance of 3-year abnormal return

(13)

13 Graph III: The cumulative average abnormal returns as a percentage (Y-axis) for a portfolio of 86 IPOs in 2012 for 36 months (X-axis)of performance after the IPO during the time-window 2007-2015. Adjustments are made for the general market (MSCI excl. US & Canada) and the financial market (iShares Global Financials). The asterisk in the graphs show whether the ARt was significant at

a 10% for that month.

The largest downturn for the MSCI abnormal returns is in month 17 with an abnormal return of -4.16% giving a t-statistic of -2.6801 which is significant. This results in the lowest cumulative abnormal return (1.61%) during the 3-year period with a t-statistic of 0.2484. During this month there are two outliers in terms of negative abnormal returns of 50% and -34%. Yet, the abnormal return remains negative with a value of -3.20% as does the linked t-statistic (-2.227) when removing these two outliers. The same procedure is used for the 30th month which has the highest positive abnormal return of approximately 4.5%. Two outliers of 51% and 31% are removed which lowers the abnormal return to 3.6% and a t-statistic of 3.3253 which is still significant the results appear to be robust.

(14)

14

the end of the 3-year period. This can be explained by the effect of the financial crisis which might have been more damaging for financial companies than for non-financial companies. The sample of 86 IPOs outperformed both markets indices.

Table VIII (appendix) shows the abnormal returns for the European and US IPOs of which the returns have been abnormal for both size and area. Table IX (appendix) shows the abnormal returns for the Asian and Latin American IPOs of which the returns have also been abnormal for size and area. The CARs of the 3-year period turn out to be not significant for all areas. There are some individual cases where the monthly ARt turns out to be significant but this does not result in a total CAR based t-statistic that is significantly different than zero. The CARs are plotted per area in Graph IV. The Latin American values turn around the most as they reach their bottom value of -37.94% in the 13th month after the IPO to increase to a maximum value of 44.29% in the 32th month and ending at 37.84% higher than the benchmark index. Out of the 36 ARt observations, 5 are significantly different than zero. Graph IV: The cumulative average abnormal returns as a percentage (Y-axis) for several portfolios based on area in 2007-2012 for 36 months (X-axis) of performance after the IPO during the time-window 2007-2015. US IPOs are abnormal for large size (S&P Banks index) and small size (S&P Regional Banks index), European IPOs for large size (Stoxxx Eur) and small size (Lyxor MSCI EMU Small Cap), Asian IPOs for large size (MSCI Pacific ex Japan) and small size (SPDR International Small Cap), Latin American IPOs for large size (MSCI Brazil Capped) and small size (SPDR International Small Cap). The asterisk in the graphs show whether the ARt was significant at a 10% for

(15)

15 -15.00% -10.00% -5.00% 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 Small Large

The performance of the US IPOs is positive at the end of the 3-year period and seems to be quite stable with only one significant ARt t-statistic out of the 36 months. None of the monthly European ARt t-statistics turns out to be significant. The abnormal returns come close to zero again after the 3-year period as they have a CAR of 1.39%. The CARs for Asia initially have a positive value but decrease over time and the value after the 3year period is -17.91% compared to the used benchmark indices. Out of the 36 ARt observations in Asia, it turns out that 7 (of which 4 are positive) of them are significantly different than zero. This makes the Asian IPOs the ones with the most significant ARt observations.

Table X (appendix) contains the returns of a split of the total sample, in small and large banks, of the IPOs abnormal for both size and area. The t-statistic of the CARs turn out not to be significant after the 3-year period for small or large banks. This implies that there is no significant size effect detected in this sample even though the literature predicted this. The t-statistic for the CAR value after 36 months of the small size banks turns out to be positive (0.7313) while the value for the large banks turns out to be even more positive (1.2090) which is not in line with other research. This implies that large banks perform better in comparison to the small banks, abnormal for the related benchmarks as can be seen in Graph V.

(16)

16

2) Stock performance using BHAR

Table XI shows the 3-years buy and hold returns abnormal for the related benchmarks. The average return of the financial market abnormal returns is higher than that of the general market abnormal returns. The t-statistic of the financial market abnormal returns is significant at a 1% level (2.8789) while the BHAR t-statistic for the general market abnormal returns is significant at a 10% level (1.6890). The 3-year holding period of the IPOs in the sample significantly outperform the financial market index. The BHAR method is geometric which can result in different values when compared to the arithmetic CAR method. In general, the BHAR method will produce less extreme and lower returns than the CAR method which is also the case for this sample as the CAR for the general market is 2.1868 versus a BHAR of 1.6890 and the CAR for the financial market is 3.3402 versus a BHAR of 2.8789. Nevertheless, the BHAR results indicate that the IPOs outperformed the market.

(17)

17 Table XI: Buy and Hold returns

Average return % Std. Dev % N t-statistic

General Market 0.1054 0.5785 86 1.6890 Financial Market 0.1851 0.5962 86 2.8789 US 0.1263 0.4840 25 1.3051 EU 0.0997 0.7157 24 0.6827 Asia -0.2167 0.5900 13 -1.3245 Latam 0.1057 0.4566 13 0.8350 Small size 0.0333 0.6195 38 0.3316 Large size 0.0605 0.5798 37 0.6349

This table presents the Buy and Hold returns (BHAR) in % for a holding period of 3-years adjusted for several indices (benchmarks). The average return of the abnormal returns is presented in the first column, the second column shows the standard deviation of these abnormal returns, the third column shows the maximum amount of observations that are used in the sample and the fourth column shows the t-statistic that is calculated to determine if the average return is statistically significant from zero. The result for different benchmarks is presented per row. The general market returns are adjusted for the MSCI ex US & Canada index, the financial market returns are adjusted for the iShares Global Financials index, the US results is adjusted for the S&P Banks index (large US banks) and the S&P Regional Banks index (small US banks), the EU return is adjusted for the Stoxxx EUR index (large European banks) and the Lyxor MSCI EMU SC index (small European banks), the Asia return is adjusted for the MSCI Pacific ex Japan index (large Asian banks) and the SPDR International Small Cap index (small Asian banks), the Latam returns are adjusted for the MSCI Brazil Capped index (large Latin American banks) and the SPDR International Small Cap index (small Latin American banks). The returns of small size contain the banks in the sample that are considered as small and are adjusted for the regional based indices that are used for small banks. The returns of large size contain the banks in the sample that are considered as large and are adjusted for the regional based indices that are used for large banks.

3) OLS model

(18)

18

(19)

19

VI. Robustness checks

Some indices (Table IV) and abnormal average returns (Table V) are skewed which might have its implications for the t-tests. The average abnormal returns adjusted for the financial market (iShares Global Financials index, Table V) are skewed which can result in a different t-statistic that might not be significant when adjusting for this. Johnson (1978) suggests an adjustment to get more accurate t-statistics:

𝑡 𝑎𝑑𝑗. = 𝑡 + 𝑔 6√𝑛+ 𝑔𝑡 𝑖² 3√𝑛 (12) With: 𝑔 =_{(𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 𝑖)³}(𝑚𝑒𝑎𝑛 𝑖) ³ (13) Where t is the originally calculated t-statistic for a particular month, mean i is the mean over the total sample per month and standard deviation i is the standard deviation over the total sample per month. The results are reported in Table XIII (appendix) for the abnormal returns of the financial market. The t-statistics are in line with the ones presented in Table VII

(appendix) that are calculated following Ritter (1991) and there are no major differences. The values that are calculated with Ritter’s test come close to those calculated with Johnson’s t-test adjustment. There appears to be no significant difference between the methods in terms of t-statistics.

(20)

20 Graph VI: The cumulative average abnormal returns as a percentage (Y-axis) in 2007-2012 for 36 months (X-axis) of performance after the IPO during the time-window 2007-2015. Both samples of 146 observations and 86 observations are adjusted for the financial market index. The asterisk in the graphs show whether the ARtwas significant at a 10% for that month.

Graph VII: The cumulative average abnormal returns as a percentage (Y-axis) in 2007-2012 for 36 months (X-axis) of performance after the IPO during the time-window 2007-2015. The samples are adjusted for the MSCI World Bank index and the iShares Global Financial index. The asterisk in the graphs shows whether the ARt of that month was significant.

0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 146 obs. 86 obs. 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 MSCI World Bank

(21)

21 -10.00% 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 1 2 3 4 5 6 7 8 9 101112131415161718192021222324252627282930313233343536 2007 IPOs 2008-2012 IPOs

In addition, an extra financial index is used to determine the robustness of this research. The MSCI World Bank Index tracks the performance of several banks in 23 developed countries. The returns are adjusted for this index and the ARt, CARt are calculated per month along with the BHAR (Graph VII, Table XV, Table XVI, appendix). The MSCI World Bank abnormal returns moves identically to the iShares Global Financial index with similar significant ARt values in the 1st, 20th and 30th month while the t-statistic after 36 months is also significant (2.8982). In addition, the BHAR t-statistic is significant (3.3791) which matches the iShares Global Financial BHAR.

The crisis started during 2007 and this might cause a difference in IPO performance between the IPO year 2007 and the IPO period 2008-2012. The model (Table XII, appendix) suggests that IPOs in 2007 outperform the market while IPOs in 2008-2012 do not. The sample is split in these periods with 41 observations in 2007 and 45 in the period 2008-2012. The results are presented in Graph VIII and Table XVII (appendix). The CAR t-statistic of the IPOs of 2007 (3.8659) after 36 months turns out to be significant while that of the IPOs during 2008-2012 is not significant (0.2079). The t-statistic of the BHAR (Table XVI, appendix) is significant for IPOs of 2007 (4.7907) and not significant for IPOs of 2008-2012 (0.7119). A possible explanation for this are high negative returns of the iShares Global Financial index (Graph I, appendix) during 2008 which results in higher abnormal returns for IPOs of 2007. Moreover, a positive market momentum of this index during 2009 can also have a larger impact on the IPOs of 2007 and improve the returns.

(22)

22

An additional model, similar to the one of Table XII (appendix), is estimated as a robustness check. It is defined as:

𝑅𝑒𝑡𝑢𝑟𝑛 𝑖 = 𝑏0 + 𝑏1 𝐿𝑜𝑔 (𝑎𝑠𝑠𝑒𝑡𝑠 𝑖) + 𝑏2 𝐿𝑜𝑔 (𝑙𝑜𝑎𝑛𝑠 𝑖) + 𝑏3 𝑎𝑠𝑠𝑒𝑡 𝑔𝑟𝑜𝑤𝑡ℎ 𝑖 + 𝑏4 (𝑎𝑠𝑠𝑒𝑡 𝑔𝑟𝑜𝑤𝑡ℎ 𝑖)2_{+ 𝑏5 𝑙𝑜𝑎𝑛 𝑔𝑟𝑜𝑤𝑡ℎ 𝑖 + 𝑏6 (𝑙𝑜𝑎𝑛 𝑔𝑟𝑜𝑤𝑡ℎ 𝑖)}2₊

𝑏7 𝑝𝑟𝑜𝑣𝑖𝑠𝑖𝑜𝑛 𝑓𝑜𝑟 𝑙𝑜𝑎𝑛 𝑙𝑜𝑠𝑠𝑒𝑠 𝑖 + 𝑏8 (𝑝𝑟𝑜𝑣𝑖𝑠𝑖𝑜𝑛 𝑓𝑜𝑟 𝑙𝑜𝑎𝑛 𝑙𝑜𝑠𝑠𝑒𝑠)2_{+ 𝑏9 2007𝑖 +} 𝑏10 2008𝑖 + 𝑏11 2009𝑖 + 𝑏12 2010𝑖 + 𝑏13 2011𝑖 + 𝑒 𝑖

Return i is the sum of the monthly average abnormal return, adjusted for the iShares Global Financials index, per bank per year divided by 3, the total amount of years in the used period. Log (assets i) is the natural logarithm of the 3-year average total assets stated on the balance sheet after the IPO per individual bank. Log (loans i) is the natural logarithm of the 3-year average total loans stated on the balance sheet after the IPO per individual bank. Asset growth i is the average asset growth from t -1 (1 year before the IPO) until the end of the 3-year period per individual bank based on total assets reported on the balance sheet. Asset growth² is the squared value of asset growth. Loan growth i is the average loan growth from t-1 until the end of the 3-year period per individual bank based on total loans as reported on the balance sheet. Loan growth² is the squared value of loan growth. Provision for loan losses i is the amount of total loan loss provision as reported divided by the total amount of loans as reported in the financial statements. Provision for loan losses² is the squared value of provision for loan losses. 2007i, 2008i, 2009i, 2010i and 2011i are time related dummies with value 0 if the IPO did not take place in that year and value 1 if the IPO took place during that year.

This model is used to examine whether the assumption of a linear relationship for growth is not false. An extra squared factor for the ratios of asset growth, loan growth and provision for loan losses is included in the model. The regression results are presented in Table XVIII (appendix). There are no significant changes in comparison to the results of the linear model (Table XII, appendix). Asset growth² and loan growth² have t-statistics of (-0.7729) and (0.1213) which are not significant. Provision for loan losses with a t-statistic of (-0.2018) and provision for loan losses² (-0.1213) are also not significant. This differs from the results found with the linear model, the t-statistic of provision for loan losses (-1.8105) was significant whereas it is not for this model. The time dummy for 2007 remains the only significant time dummy at a 10% level. There appears to be no exponential function for asset growth, loan growth and provision for loan losses.

(23)

23

VII. Discussion

The results of this research are based on a sample of 86 IPOs of banks during the event window 2007-2012. This sample can be considered as rather small which can have its implications for the inferences based on the tests. In addition, as can be seen in Table II the year 2007 contains nearly half of all the IPOs of the sample. This is possibly caused by the financial crisis that started in 2007 which makes it less attractive for firms and banks in special to go public. The other years are under-represented in comparison to 2007 which makes the overall sample less reliable and makes it more difficult to get an accurate result for time effects. The same goes for the model that is based on financial and accounting data. A sample of 44 banks is used which might not be representative for the population as a whole. The sample size should be increased to get a more accurate result but this includes extending the event window as well. Furthermore, there might be a self-selection bias caused by the chosen event window. Banks that go public during a distressed time might be in such good shape both financially and operationally that they do not represent the populations of banks that would go public during normal market circumstances. This results in a self-selection of mostly banks that outperform the market which will not be generalizable for times when market momentum is different. However, the robustness check on timing of the IPOs (Graph VIII) shows that IPOs in 2007 did outperform the market while IPOs from 2008-2012 did not This suggests that the crisis did have an impact on the performance which can be used as an additional research subject and that there was no self-selection during the crisis.

(24)

24

VIII. Conclusion

This paper examines the 3-year after-IPO performance of 86 banks that went public during the period 2007-2012. It is concluded that these IPOs outperform both the worldwide general and financial market indices when applying the CAR method. The outperformance is statistically significant when applying the BHAR method for the general market but only on a 10% level whereas the outcome of this method is significant for returns that have been adjusted for the financial market at a 1% level. This difference in significance between the BHAR and CAR measure for the general market abnormal returns may be caused by the fact that the BHAR measure is a geometric method and the CAR is a arithmetic method. In addition, there is no outperformance when adjusting for size and area except for one estimation that shows underperformance for European IPOs.

There are no significant effects of asset growth and loan growth on the 3-year performance. It is expected that growth would have a negative effect on the performance since it increases the risk but this is not concluded from this research. Examining the timing effects of the IPOs results in the outcome that IPOs that took place in 2007 outperformed the market. This signals that there might be a financial crisis related effect as these IPOs have less exposure to the consequences of the financial crisis in their 36-months performance.

Literature on comparing non-financial IPOs with financial IPOs suggests that the latter group outperforms the market following the IPO. This suggestion is confirmed by this research as the bank IPOs outperform several market indices and with different measures. However, literature also suggests that higher growth will have a negative effect on the performance of banks which is different in comparison to non-banks. This effect is not found in this research which might be caused by the timing of the IPOs.

(25)

25

(26)

26

IX. References

Beatty, R., Ritter, J., 1985, Investment banking, reputation, and the underpricing of Initial Public Offerings, Journal of Financial Economics 15, 213-232

Bessler, W., Murtagh, J., Siregar, D., 2003, Dividend policy of bank Initial Public Offerings, In: Bagella, M., Bechetti, L., Hasan, I. (Ed.), Transparency, governance and markets, Amsterdam, The Netherlands, pp.187-228

Bessler, W., Nohel, T., 1996, The stock market reaction to dividend cuts and omissions by commercial banks, Journal of Banking and Finance 20, 1485-1508

Bessler, W., Nohel, T., 2000, Asymetric information, dividend reductions and contagion effects in bank stock returns, Journal of Banking and Finance 24, 1831-1848

Cornett, M., Fayman, A., Marcus, A., Tehranian, H., 2010, Dividends, maturity, and acquisitions: Evidence from a sample of bank IPOs, Review of Financial Economics 20, 11-21

Cumming, D. J., Walz, U., Werth, J.C., 2016, Entrepreneurial Spawning: Experience, education and exit, unpublished working paper, Financial Review

Fama, E.F., 1980, Banking in the theory of finance, Journal of Monetary Economics 6, 39-57 Fama, E.F., French, K.R., 1993, Common risk factors in the return on stocks and bonds, Journal of Financial Economics 25, 23-49

Francis, B., Hasan, I., Siregar, D., 2009, The choice of IPOs versus M&A: evidence from the banking industry, Applied Financial Economics 19, 1987-2007

Gande, A., Saunders, A., 2012, Are banks still special if there is a secondary market for loans?, Journal of Finance 67, 1649-1684

Gandhi, P., Lustig, H., 2015, Size anomalies in U.S. bank stock returns, Journal of Finance 70, 733-768

(27)

27

Houge, T., Loughran, T., 1999, Growth fixation and the performance of bank initial public offerings 1983-1991, Journal of Banking and Finance 23, 1277-1301

Ibbotson, R., 1975, Price performance of common stock new issues, Journal of Financial Economics 3, 235-272

Johnson, N. J., 1978 Modified t-tests and confidence intervals for asymmetrical populations, Journal of the American Statistical Association, 73, 536-544

Kao, J.L., Wu, D., Yang, Z., 2009, Regulations, earnings management, and post-IPO performance: The Chinese evidence, Journal of Banking and Finance 33, 63-76

Lee, C., Schleifer, A., Thaler, R., 1991, Investor sentiment and the closed-end fund puzzle, Journal of Finance 46, 75-109

Loughran, T., 1993, NYSE vs. NASDAQ returns: Market microstructure or the poor performance of Initial Public Offerings?, Journal of Financial Economics 33, 241-260

Loughran,, T., Ritter, R., 1995, The new issue puzzle, Journal of Finance 50, 23-51

(28)

28

X. Appendix

Table I: the total amount of countries in the sample calculated per individual country

Country Count

Spain 3

Italy 6

Russia 1

Germany 2

United States of America 25

Switzerland 3 Republic of Korea 2 Denmark 3 United Kingdom 2 Japan 3 Poland 2

Syrian Arab Republic 3

Norway 2 Colombia 1 Brazil 10 Finland 1 Turkey 1 Qatar 1 Mexico 2 India 4 Saudi Arabia 1 China 6 Vietnam 2

Table III: The total amount of countries in a particular area. The grouping is as follows: US contains United States of America, Asia contains Republic of Korea, Japan, China and Vietnam, Europe contains Spain, Italy, Germany, Switzerland, Denmark, United Kingdom, Poland, Netherlands, Norway, Finland and Austria, Latin America contains Colombia, Brazil and Mexico

Area Count

US 25

Asia 13

Europe 24

(29)

29 Table VI: In the left part of the table, the amount of IPOs per area are presented for the sample of 44 banks. In the right part of the table, the amount of IPOs per year are reported for the sample of 44 banks.

Area Count Year Count

(30)

30 Table VII: the average firm-abnormal returns (ARt) and the cumulative average returns (CAR) in

percentage per month after going public with t=1 as the return at the end of the first month that a company is public. The t-statistics of both the ARt and CAR are presented, these are calculated with

formulas (4), (5) and (6). The left side of the table represents the results adjusted for the general market index (MSCI excluding US & Canada) and the right side of the table represents the results adjusted for the financial market index (iShares Global Financials).

Month General market Financial market

ARt % T-stat CAR % T-stat ARt % T-stat CAR % T-stat

(31)

31 Table VIII: the average firm- abnormal returns (ARt) and the cumulative average returns (CAR) in

formulas (4), (5) and (6). The left side of the table represents the results adjusted for the US market indices (S&P Banks for large US banks and S&P Regional Banks for small US banks) and the right side of the table represents the results adjusted for the European market indices (Stoxxx Eur for large European banks and Lyxor MSCI EMU Small Cap for small European Banks).

Month US EU

(32)

32 Table IX: the average firm- abnormal returns (ARt) and the cumulative average returns (CAR) in

formulas (4), (5) and (6). The left side of the table represents the results adjusted for the Asian market indices (MSCI Pacific ex Japan for large Asian banks and SPDR International Small Cap for small Asian banks ) and the right side of the table represents the results adjusted for the Latin American market indices (MSCI Brazil Capped for large Latin American banks and SPDR International Small Cap for small Latin American banks).

Month Asia Latin America

(33)

33 Table X: the average firm- abnormal returns (ARt) and the cumulative average returns (CAR) in

formulas (4), (5) and (6). The left side of the table represents the results adjusted for the small market indices (S&P Regional Banks for US, Lyxor MSCI EMU Small Cap for Europe, SPDR International Small Cap for Asia and Latin America). The right side of the table represents the results adjusted for the large market indices (S&P Banks for US, Stoxxx Eur for Europe, MSCI Pacific ex Japan for Asia, MSCI Brazil Capped for Latin America).

Month Small size Large size

(34)

34

Table XII: the results of the ordinary least squares regression model: Returni = b0 + b1 Log (assetsi) + b2 Log (loansi) + b3 asset growthi + b4 loan growthi + b5 provision for loan lossesi + b6 EUi + b7 USi + b8 Asiai + b9 2007i + b10 2008i + b11 2009i + b12 2010i + b13 2011i + e i. Returni is the sum of the monthly average abnormal return, adjusted for the iShares Global Financials index, per bank per year divided by 3, the total amount of years in the used period. Log (assetsi) is the natural logarithm of the 3-year average total assets stated on the balance sheet after the IPO per individual bank. Log (loansi) is the natural logarithm of the 3-year average total loans stated on the balance sheet after the IPO per individual bank. Asset growth i is the average asset growth from t -1 (1 year before the IPO) until the end of the 3-year period per individual bank based on total assets reported on the balance sheet. Loan growth i is the average loan growth from t-1 until the end of the 3-year period per individual bank based on total loans as reported on the balance sheet. Provision for loan losses

i is the amount of total loan loss provision as reported divided by the total amount of loans as reported in the

financial statements. EUi, USi, and Asiai, are dummies for the areas in which the IPO took place with value 0 if the IPO was not in that area and value 1 if it was in that area. 2007i, 2008i, 2009i, 2010i and 2011i are time related dummies with value 0 if the IPO did not take place in that year and value 1 if the IPO took place during that year. The first column reports the variables. The second column reports the coefficient and t-statistic in brackets related to the variable. In addition, the total number of observations, F-statistic, R-squared and Adjusted R-squared is reported.

Variable Coefficient (t-statistic) Constant 0.0999 (0.3673) Log assets 0.0171 (0.0905) Log Loans -0.0118 (-0.0613) Asset growth 0.3667 (0.8094) Loan growth -0.3616 (-1.0678)

Provision for loan losses -3.7211

(-1.8105) EU -0.2344 (-1.9635) US -0.1059 (-0.9153) Asia -0.1641 (-1.4450) 2007 0.2245 (1.7448) 2008 0.167 (1.2403) 2009 -0.0439 (-0.3340) 2010 0.0919 (0.6355) 2011 0.1438 (1.1577)

Number of observations 44 R-squared 0.4523

F-statistic

(35)

35 Table XIII: the average firm-abnormal returns (ARt) and the cumulative average returns (CAR) in

formulas (4), (5) and (6) followed by the adjustment of Johnson (1978) with the formulas (12) and (13). The returns are adjusted for the financial market (iShares Global Financials). The left table shows the values calculated with Johnson’s adjustment, the right side presents the originally calculated values with Ritter’s method.

Month Financial market adjusted Financial market original

(36)

36 Table XIV: the average firm-abnormal returns (ARt) and the cumulative average returns (CAR) in

formulas (4), (5) and (6). The returns are adjusted for the financial market (iShares Global Financials). The left table shows the values calculated with a sample of 149 IPOs, the right side presents the originally calculated values with 86 IPOs.

Month Financial market 149 obs. Financial market original 86 obs.

(37)

37 Table XV: the average firm-abnormal returns (ARt) and the cumulative average returns (CAR) in

formulas (4), (5) and (6). The returns are adjusted for the added financial market (MSCI World Bank index) in the left table and the originally used financial market (iShares Global Financial) in the right table.

Month Financial market MSCI World Bank index Financial market original iShares Global

(38)

38 Table XVI: Buy and Hold returns of the robustness checks

Average return % Std. Dev % N t-statistic

MSCI World Bank 0.2133 0.5854 86 3.3791

2007 0.3058 0.4086 41 4.7907

2008-2012 0.0752 0.7084 45 0.7119

(39)

39 Table XVII: the average firm-abnormal returns (ARt) and the cumulative average returns (CAR) in

formulas (4), (5) and (6). The returns are adjusted for the financial market (iShares Global Financials). The left table shows the values calculated with a sample of 40 IPOs that took place in 2007, the right side presents the calculated values with 46 IPOs that took place during the period 2008-2012.

Month Financial market 2007 Financial market 2008-2012

(40)

40

Table XVIII: the results of the ordinary least squares regression model: Returni = b0 + b1 Log (assets)i + b2 Log (loans)i + b3 asset growthi + b4 (asset growth)² + b5 loan growthi + b6 (loan growth)² + b7 provision for loan lossesi + b8 (provision for loan lossesi)² + b9 2007i + b10 2008i + b11 2009i + b12 2010i + b13 2011i + ei. Returni is the sum of the monthly average abnormal return, adjusted for the iShares Global Financials index, per bank per year divided by 3, the total amount of years in the used period. Log (assetsi) is the natural

logarithm of the 3-year average total assets stated on the balance sheet after the IPO per individual bank. Log (loansi) is the natural logarithm of the 3-year average total loans stated on the balance sheet after the IPO per individual bank. Asset growth i is the average asset growth from t -1 (1 year before the IPO) until the end of the 3-year period per individual bank based on total assets reported on the balance sheet. Asset growth² is the squared value of asset growth. Loan growth i is the average loan growth from t-1 until the end of the 3-year period per individual bank based on total loans as reported on the balance sheet. Loan growth² is the squared value of loan growth. Provision for loan losses i is the amount of total loan loss provision as reported divided by the total amount of loans as reported in the financial statements. Provision for loan losses² is the squared value of provision for loan losses. 2007i, 2008i, 2009i, 2010i and 2011i are time related dummies with value 0 if the IPO did not take place in that year and value 1 if the IPO took place during that year. The first column reports the variables. The second column reports the coefficient and t-statistic in brackets related to the variable. In addition, the total number of observations, F-statistic, R-squared and Adjusted R-squared is reported.

Variable Coefficient (t-statistic) Constant 0.0314 (0.1510) Log assets 0.1453 (0.7716) Log Loans -0.1663 (-0.0613) Asset growth 0.9713 (0.8355) Asset growth² -1.2981 (-0.7729) Loan growth -0.0482 (0.0481) Loan growth² -0.2491 (-0.1213)

Provision for loan losses -1.3151

(-0.2018)

Provision for loan losses² 10.6481

(-0.1213) 2007 0.2589 (2.0218) 2008 0.1616 (1.1595) 2009 -0.0548 (-0.4372) 2010 0.1371 (0.8702) 2011 0.1035 (0.8784)

Number of observations 44 R-squared 0.4355

F-statistic

(41)

41 Graph I: The abnormal returns of the MSCI excluding Canada and the U.S. index and the iShares Global Financial EFT per observation (monthly)