Are banks more opaque? Evidence from Insider Trading 1

Are banks more opaque? Evidence from Insider Trading 1 Fabrizio Spargoli a and Christian Upper b a Rotterdam School of Management, Erasmus University b Bank for International Settlements Abstract We investigate whether banks are more opaque than other firms using data on trades by insiders. Our findings indicate that purchases by bank insiders earn a higher return than purchases by other firms insiders only at short horizons. By contrast, the return on sales by bank insiders is consistently greater over all the horizons up to 180 days. Returns on insider trading do not substantially change during times of crisis. Moreover, the size of the loan book and loan loss allowance are the main determinants of the cross sectional variation in the returns on purchases and sales by bank insiders, respectively. Overall, our findings suggest a greater opacity of banks especially with regards to negative information. Key words: Bank opacity, insider trading, financial stability JEL classification: G14, G20, G21 1 Corresponding author: Fabrizio Spargoli; Rotterdam School of Management, Erasmus University. Burgemeester Oudlaan 50, 3062 PO Rotterdam. Email: spargoli@rsm.nl

1. Introduction It is conventional wisdom that the degree of asymmetric information between insiders and outside investors is higher for banks than other firms. According to Morgan (2002), this argument hinges on three reasons. First, there is limited public information on the typical bank asset, that is loans. There is no market for loans, and hence no price, and banks often have soft, non quantifiable information on borrowers. Second, any information disclosure by banks might quickly become out of date, as some assets can be traded very quickly in liquid markets. Third, high leverage may lead to significant agency problems. For example, banks might tilt portfolio towards lower-yielding opaque assets to escape market discipline (Wagner, 2007). Hence, according to the conventional wisdom, it is in the nature of banks to be more opaque than other firms. Since opacity prevents an effective discipline by outside investors, the logical implication is that banks should be regulated more than any other firm. However, the arguments why banks are opaque could apply also to firms in other sectors. Oil firms, for example, own reserves in the ground. There is no market for these reserves, and their size, as well as the costs to extract them, are often difficult to assess for an outsider. Another example are technology firms. Typically, these firms make large investments in research and development projects. The outcome of these projects are often unique to a firm and are not traded in a market. Moreover, accounting rules consider research and development expenditures as cost items deductible from profits on a yearly basis. This implies it is not possible to track changes in the value of research and development projects, in contrast to other assets valued at market prices on the balance sheet (Aboody and Lev, 2001). These characteristics make technology firms difficult to value for an outsider, to a not necessarily lower extent than banks. Hence, whether banks are more opaque than other firms is ultimately an empirical question. This paper uses trades by insiders in US firms to investigate whether banks are more opaque than firms in other sectors. The 1934 Securities and Exchange Act defines insiders as corporate officers, directors, and owners of 10 percent or more of any equity class of securities. Moreover, this act requires insiders to publicly disclose information on their trades, including the type of transaction, size, and execution price. Using these data, as well as information on stock returns, we compare the return and the profit on insiders' trades at different time horizons for banks and non banks. The logic of our argument is that superior information about an asset represents a source of monopoly power, which can be exploited by

trading that asset. Since collecting information is costly, the equilibrium trade and price will be such that the marginal gain and cost of information are equal. Compared to outsiders, insiders presumably have a lower cost of collecting information, because of their position in the firm. This cost advantage should give insiders a higher monopoly power than outsiders, regardless of the type of firm. However, if outsiders find it more difficult to value banks than other firms, bank insiders should enjoy greater market power than insiders in other firms. Hence, the hypothesis of a greater opacity of banks should be reflected in higher profits from insider trading in banks compared to other firms. We find that, compared to insiders in other firms, bank insiders earn a higher return on purchases, and a lower return on sales, only in the period between the trade and its disclosure. At longer horizons, there is no significant difference in returns on purchases, whereas the return on sales by bank insiders is lower than other firms. This evidence, which is confirmed when using profits instead of returns on trades, suggests an advantage of bank insiders especially with respect to negative information. Hence, our findings have implications for financial stability, since they suggest the inability of markets to discipline banks. A corollary to the argument why banks are more opaque is that information should be more valuable in times of stress. When volatility is high, it is more difficult to value an asset, because existing information is of little use for an outsider. Moreover, in times of stress, the distribution of returns on assets becomes more widespread, especially on the left tail. These two arguments imply a higher return on private information in times of stress, as well as a higher cost of collecting this information for outsiders. Hence, insiders should enjoy a greater market power in times of crisis, because of their greater cost advantage compared to normal times. Empirically, we should observe that the profits from insider trading in banks, compared to firms in other sectors, should be even larger in times of stress. Considering the 2007-09 financial crisis as a period of stress, we do not find consistent evidence of greater gains by bank insiders. Returns do not vary significantly during crises times, while profits on sales by bank insiders increase, but only for the longest horizons (90 and 180 days). Our main finding is that the results obtained pooling the time series of returns are driven mainly by differences between banks and non-banks during normal times. Having compared the return and profit on trades by insiders in banks and non banks, we proceed by investigating the cross-sectional variation within the banking sector. We relate the returns on insider trades in banks to a number of balance sheet ratios, with the aim to point

out the determinants of bank opacity. Our results suggest a crucial role of traditional banking activities. Banks with a higher loan to equity ratio exhibit greater returns on purchases, whereas the returns on sales decrease with the loan loss allowance to equity ratio. Other factors, such as leverage, real estate and opaque assets, are associated with returns on insider trades, but their economic significance is smaller. Finally, we highlight how the degree of bank opacity does not seem to be related to investment banking activities. Our paper is not the first one trying to establish whether banks are more opaque than other firms. Morgan (2002) shows that banks are more likely to have split ratings than other firms, suggesting that they are more opaque. By contrast, Flannery et al. (2004) find that bank stocks have similar bid-ask spreads and price impact of trades as the stocks of other firms, indicating a similar degree of opacity. A subsequent paper by the same authors (Flannery et al, 2013) confirms this results, but documents larger spreads and price impact measures during financial crises. Dewally and Shao (2013) reveal a positive link between the use of financial derivatives and the opacity of banks, as measured by the correlation between stock price of banks and the market index. Our contribution to this literature is to use an alternative measure of asymmetric information, that is profits from insider trading. Collin-Dufrense and Vos (2015) show that the standard measures of liquidity and adverse selection used in the market microstructure literature, such as bid-ask spread and price impact of trades, do not capture informed trading. Activist investors, namely investors with an informational advantage compared to outsiders, accumulate trades in periods when markets are liquid and adverse selection is low, according to the standard measures. In the light of these findings, it is useful to provide new evidence on whether banks are more opaque than other firms using returns and profits from insider trading. There exists a large literature investigating whether insider traders earn abnormal profits, 2 but few papers investigate the across firm variation of this returns. Aboody and Lev (2001) focus on RD expenditures, and show that insiders in RD firms earn a higher return. Lakonishok and Lee (2001) document that trades by insiders predict stock returns, especially for small firms. We contribute to this literature by comparing banks to firms in other sectors, and showing that bank insiders obtain a higher return. Moreover, we investigate the variation across banks, relating the returns by bank insiders to bank balance sheet characteristics. 2 See, for example, Seyhun (1986), Seyhun (1992), and Jeng, Metrick, and Zeckhuser (2003).

Other papers used insider trading data to examine whether bank insiders foresaw the 2007-09 financial crisis. Fahlenbrach and Stultz (2011) do not find evidence of managers reducing their holding of shares before or during the crisis, which is against the hypothesis of a conscious excessive risk taking leading to the 2007-09 crisis. Adebambo et al (2015) reach a similar conclusion comparing net purchases by managers of banks and non financial firms. In contrast, distinguishing banks based on their exposure to the housing market and using insider trading data from 2006, Cziraki (2015) finds suggestive evidence that bank managers were able to foresee the financial crisis. Differently from these studies, our paper aims to assess whether bank insiders earned higher profits during the recent financial crisis, abstracting from their predictive ability. This paper is structured as follows. Section 2 describes the dataset and presents the summary statistics. Section 3 and 4 contain the main analysis on the difference in insider gains between banks and other firms, using returns and profits. Section 5 examines the market reaction to the disclosure of insider trades, while Section 6 investigates the effect of the financial crisis of 2007-09. Section 7 documents the determinants of the returns on trades by bank insiders, and Section 8 concludes. 2. Data Data on insider trade come from Thomson Reuters, and refer to Form 3, 4, and 5 transactions. We focus on purchases and sales, excluding other types of transactions such as the exercise of options, grants and awards. Following Lakonishok and Lee (2001), we focus on trades by CEOs, CFOs, chairmen of the board, presidents, directors, officers, and vice-presidents. We exclude transactions of less than 100 shares, those for which the price is not within 20% of the CRSP closing price on the same day, and those for which the number of shares traded is greater than 20% the number of shares outstanding. We collect information on prices, returns, shares outstanding and trading volumes information from CRSP. Following Flannery et al (2004, 2012), we define as banks the firms with SIC codes 6021-6025 and 6710-6712. We match the CRSP dataset with the one on insider trades using the 8-digit CUSIP codes. Moreover, we also match on historical CUSIP codes (NCUSIP), to make sure we are not excluding firms whose identifier changed during our sample period. Finally, we obtain data on bank balance sheets from the Call reports filed with the FDIC. We merge this dataset with the CRSP and insider trades dataset using the linking tables provided

by the Federal Reserve Bank of New York (2014). 3 These tables link the firm identifier in the Call Reports (RSSID) to one of the firm identifiers in CRSP (PERMCO). We obtain balance sheet information for a total of 5681 quarter-bank observations. 3. Buy and Hold Returns from Insider Trading: Banks vs. Non-Banks Our main empirical test is based on comparing the gains from trade by insiders in banks and other firms. To accomplish this, we start by considering buy-and-hold returns, defined as the average return of the stock over a certain time horizon starting from the trading date. We consider 15, 30, 60, 90, and 180 days from the trading date and, following Aboody and Lev (2001), the period between the trading and the disclosure date, which is on average 11 days (the median is 2 days). This interval is particularly relevant for our analysis, because trades by insiders are unknown to the other market participants. Hence, insiders can exploit most of the rents from their private information. Our empirical test is analogous to Aboody and Lev (2001). We construct monthly portfolios of stocks based on two criteria. The first is whether the firm is a bank, as defined as in Flannery et al. (2004), or another type of firm. The second is whether, in a given month, insiders were net purchasers or sellers of the stock. With this procedure, we obtain four portfolios of stocks. We calculate the return of these portfolios as the (non weighted) average of the buy-and-hold returns of the individual stocks, based on their trading date. For example, the return on the portfolio of banks for which insiders are net sellers in September 2010 is calculated as the monthly average of the return on the stocks in that portfolio that were traded in that month and year. We estimate the following model: Our dependent variable is the difference in the buy-and-hold return on the bank and non bank portfolio at time t, in the months where insiders were net purchasers (NP equal to one) or sellers of shares. The independent variables are the three Fama-French factors. The first is the market return at time t, net of the risk free interest rate. The second factor is size, defined as the difference in the return of a value weighted portfolio of small and large stocks at time t. 3 https://www.newyorkfed.org/research/banking_research/datasets.html

Finally, the third factor is book-to-market, that is the difference in the return of a value weighted portfolio of high and low book-to-market stocks at time t. We are interested in the value of the intercept, which captures the risk-adjusted difference in the returns of the bank and non-bank portfolio. If banks are more opaque than other firms, insiders should obtain higher returns in months where they are net buyers ( ), and lower returns in months where they are net sellers ( ). Table 2 contains the results. Panel A shows that, in months where insiders are net buyers, the return on the bank portfolio is significantly higher only in the time period between the trading and disclosure date. The coefficient is statistically significant at the one percent level and indicates that, in months where insiders are net buyers, the return on the bank portfolio is more than double (factor of 2.5) the sample average return on the non-bank portfolio. Panel B contains the results for the months where insiders were net sellers. The return on the bank portfolio is significantly lower than the non bank portfolio at all horizons, excluding the 30 and 180 days. The coefficient is higher (in absolute value) at shorter horizons, but p- values are lower at longer horizons. As for their economic significance, our estimates indicate that the return on the bank portfolio is 50% lower than the non bank portfolio at the 15 days horizon, and 24% lower at the 90 days horizon. Overall, the evidence on the buy-and-hold returns on insider trades suggest that bank insiders are more informed than non bank insiders especially when they sell their stocks. Following a purchase, bank insiders obtain a higher return only in the short term. 4. Profits from Insider Trading: Banks vs. Non-Banks In this Section, we estimate equation (1) using insiders profits as a dependent variable. This measure is calculated as the product of the cumulative buy-and-hold return and the size of the trade, considering avoided losses as profits. The trading profit is a relevant measure of insiders' information advantage, because of two reasons. First, market microstructure models postulate traders with the objective to maximize profits. Second, profits are calculated using the size of the trade, which might reflect insider information as well as strategic considerations on the price impact of a trade.

Table 3 contains the results. Panel A shows that, in months where insiders are net buyers, the average profit on the bank portfolio is never significantly higher than the non-bank portfolio, regardless of the time horizon. In fact, trading profits are significantly smaller at all time horizons, except for the trade-disclosure interval. The difference in trading profit is the lowest for the 15 and 30 days horizon. The coefficient indicates that, in months where insiders are net buyers, the 30 days profit on the bank portfolio is 70% smaller the sample average profit on the non-bank portfolio. Panel B contains the results for the months where insiders were net sellers. The average profit on the bank portfolio is larger than the non bank portfolio at all horizons, and statistically significant at least at the 5 percent level. The difference is higher (in absolute value) at shorter horizons. At the 15 day interval, our estimates indicate that the average profit on the bank portfolio is double the profit on the non bank portfolio. At the longest horizon, insider traders selling bank stocks earn a 80% higher profit than the insiders selling other stocks. Overall, the evidence on profits reinforces the one on buy-and-hold returns on insider sales, but weakens even more the one on insider purchases. In months where insiders are net sellers, the return and, even more so, the profit on the bank portfolio are greater than in the non bank portfolio. By contrast, in months where insiders are net buyers, the return on the bank portfolio is higher only at the shortest horizon, whereas profits are the same as for non banks, if not lower. To summarize, the evidence on returns and profits suggests that bank insiders have an informational advantage when they sell their stocks but not when they buy. This result highlights the importance of the type of information in assessing the opacity of banks. Opacity is more severe for banks when information is bad, namely the scenario that regulators should be more concerned about. 5. Investors Reaction to the Disclosure of Trades This section investigates the market reaction to the disclosure of insider trades. If these trades contain relevant private information, and banks are more opaque than non banks, the postdisclosure return should be greater (smaller) for banks conditionally on insider purchases (sales).

To test this hypothesis, we follow Aboody and Lev (2001) and compare the raw returns of banks and non banks at 1, 2 and 3 days after disclosure. In contrast to Section 3 and 4, we do not consider monthly portfolios, as we are interested in the market reaction to disclosure. Our strategy is to regress the raw returns after disclosure on a dummy equal to one for banks and zero for other firms, splitting the sample into insider sales and purchases. Since our analysis is based on individual trade data, we do not use net purchases, but whether each trade is a purchase or sale. Table 4 contains the results. The bank dummy is negative and statistically significant at least at the 5-percent level when using the returns at 1 and 2 days after disclosure and conditioning on sales. By contrast, there is no significant difference in the returns of banks and other firms after the disclosure of purchase trades. Overall, these results suggest that the market reacts more to the disclosure of sale trades by banks. By contrast, there is no difference between banks and non banks in the market reaction following the disclosure of a purchase trade. This result suggest that sales by bank insiders are more informed than purchases, confirming the findings in Section 3 and 4. 6. Insider Gains in Times of Crisis The findings in the previous sections suggest that banks, on average, are more opaque than other firms, especially with regards to negative information. In this section, we investigate whether there is variation over time in this findings, focusing on the financial crisis of 2007-2009. One of the factors leading to this event, according to a widespread view, was the opacity of banks, which impaired the ability of outside investors to assess bank solvency. Hence, based on this view of the crisis, we should expect higher buy-and-hold returns and profits on trades by bank insiders vis-à-vis other firms. To test for differences in insider gains during the financial crisis of 2007-2009, we augment equation (1) with a crisis dummy ( ) equal to one from August 2007 until September 2009. Hence, this dummy captures whether the difference in the return on trading the bank and non-bank portfolio during the crisis was smaller or greater than in normal times. If banks are more opaque than other firms, we should observe a positive (negative) coefficient on the crisis dummy in the months where insiders are net buyers (sellers).

Table 5 contains the results using buy-and-hold returns as dependent variable. In the months where insiders are net buyers (Panel A), the crisis dummy is positive and statistically significant at the 5 percent level only at the 30 days horizon. The coefficient indicates that the difference in the return on banks and non-banks was larger by almost 2 basis points during the financial crisis of 2007-09. Comparing these findings to the averages over the whole time period (Table 3, Panel A), it is worth noting that the constant remains positive and statistically significant at the trade-disclosure horizon. Hence, in the short term, the higher returns on purchases by bank insiders do not depend on more severe information asymmetries during times of crisis. Panel B of Table 5 illustrates the results using buy-and-hold returns as dependent variable and conditioning on months where insiders are net sellers. The crisis dummy is negative and statistically significant at the 1-percent level only at the trade-disclosure horizon. The coefficient indicates that, compared to normal times, the difference in the return on the bank and non-bank portfolio was smaller by almost 30 basis points during the crisis of 2007-09. Moreover, the crisis dummy completely absorbs the average effect showed in Panel B of Table 3. This suggests that, at the trade-disclosure horizon, bank insiders have an information advantage mainly in times of crisis. By contrast, at longer horizons, the difference in the buyand-hold returns on the bank and non-bank portfolio does not change during times of crisis. Moreover, except for the 15 days horizon, the informational advantage by bank insiders uncovered in Panel B of Table 3 is explained by normal times rather than crises. The evidence on profits, which is reported in Table 6, is in line with the one on returns, except for a few differences. First, bank insiders do not obtain higher profits purchasing stocks in times of crisis, regardless of the time horizon. In fact, the crisis dummy is negative and statistically significant at longer horizons, suggesting lower profits on purchases by bank insiders. By contrast, in months where insiders are net sellers, profits are larger in times of crisis at the 90 and 180 days horizons. Moreover, at all but the 30 days horizon, bank insiders obtain a higher profit than non-bank insiders when selling their stocks in normal times. Overall, these findings provide limited evidence in favour of the hypothesis that the gains from insider trading in banks are larger in times of crisis. Returns on purchases are higher only at the 30 days horizon, whereas returns on sales are smaller only at the trade-disclosure horizon. As for profits, the main finding is that bank insiders selling their stock are better off in times of crisis at the longest horizons. Moreover, controlling for the financial crisis of

2007-09 does not substantially alter the main findings based on the entire sample. This implies that the greater advantage of insiders especially with regards to negative information seems to be a typical difference between banks and other firms, which cannot only be attributed to turbulent times. Finally, our findings contrast with Flannery et al (2012), who finds evidence of a greater opacity of banks in crisis times. To reconcile these findings with ours, let us highlight the insights from Collin-Dufresne and Vos (2015). These authors document that insiders trade more intensely when measure of market liquidity, such as those used in Flannery et al (2013), are high. Hence, the evidence on bid-ask spreads and price impact of trades in Flannery et al (2013) does not necessarily imply that asymmetric information between insiders and outside investors was more severe during the 2007-09 financial crisis. 7. The Determinants of Insider Gains in Banks Having compared the gains from trade by bank and non-bank insiders, we now focus on the cross-section of banks. Our aim is to investigate the bank-level determinants of insider gains. We are interested in the following balance sheet ratios: Total loans, net of the allowance for loan and lease losses, normalized by the market value of equity; Loan loss allowance, normalized by the market value of equity; Fair value of assets held in trading accounts, normalized by the market value of equity; Other real estate owned, normalized by the market value of equity. This balance sheet category primarily includes real estate taken in settlement of problem loans, though some real estate investments (other than bank premises) are also included; Opaque assets, normalized by the market value of equity. Following the definition in Flannery et al (2004), this variable is the sum of the book value of bank premises and fixed assets, investments in unconsolidated subsidiaries, intangible assets, and the balance sheet category other assets ; Non interest rate income, normalized by the sum of interest and non interest rate income. This is a measure of diversification into non traditional banking activities; Net income, normalized by the market value of equity;

Market leverage, calculated as the sum of the book value of liabilities and market value of equity, divided by the market value of equity. Due to the time variation of the bank-level information from the Call Reports, we average the buy-and-hold returns by bank insiders at the bank-quarter level. To account for the fact opacity increases (decreases) with returns in case of purchases (sales), we split the sample based on whether the total amount of purchases of a stock, net of sales, is positive in a certain quarter. For each subsample, we estimate the following regression model: The dependent variable,, is the average buy-and-hold return on the stock i traded during quarter t. Our variables of interest are those in the set, which includes the previously mentioned balance sheet ratios. We control for bank fixed effects,, and a set of control variables,. This set includes the quarterly average market value of equity (in logs), and the inverse of the average quarterly share price, as in Flannery (2004). Table 7 contains the results, using the buy-and-hold returns on trades by bank insiders at different horizons as dependent variable. In months where insiders are net buyers (Panel A), banks with a higher amount of loans and assets in trading accounts, a greater non interest rate income ratio and a lower leverage exhibit higher buy-and-hold returns at medium-long horizons (60, 90, and 180 days). At shorter horizons, namely during the trade to disclosure interval, the buy-and-hold returns on purchases by bank insiders are mainly driven by the amount of real estate assets, leverage and profitability. In terms of economic significance, the stronger association is with the loan to equity and leverage ratios. A standard deviation increase in the former (latter) ratio corresponds to a 193 (160) standard deviation increase (reduction) in the buy-and-hold return at 60 days. For the other horizons, the orders of magnitude is similar. In months where insiders are net sellers (Panel B), banks with a higher amount of real estate assets and loan loss allowance exhibit lower buy-and-hold returns at long horizons (90 and 180 days). At the trade-disclosure horizon, buy-and-hold returns decrease with the size of opaque assets and increase with profits. In terms of economic significance, the stronger association is with loan loss allowances and opaque assets. A standard deviation increase in the loan loss allowance to equity ratio corresponds to a 130 standard deviation reduction in

the buy-and-hold return at 90 days, with a similar order of magnitude at the 180 days horizon. As for the opaque assets to equity ratio, a standard deviation increase in this ratio is associated with a 102 standard deviations reduction in the buy-and-hold return at the tradedisclosure horizon. Overall, the results in Table 7 suggest that the determinants of buy-and-hold returns differ depending on the horizon and whether insiders are net sellers or buyers of bank stocks. Factors related to the typical bank activities, such as the size of the loan portfolio and the loan loss allowance, matter for long run returns. By contrast, short run returns are mainly driven by profitability and the value of other assets than loans, such as intangible assets and real estate. Moreover, the size of the loan portfolio and loan loss allowances, that are factors related to the typical bank activities, have the strongest impact on the returns in months where insiders are net sellers and buyers, respectively. 8. Conclusions This paper examines whether banks are more opaque than other firms using data on trades by firm insiders. Comparing the returns and the profits on a portfolio of banks and non-banks traded by insiders, we find evidence suggesting a higher opacity of banks especially with regards to negative information. The difference in the return and profit on the bank and nonbank portfolios is negative in months were insiders are net sellers, from the shortest (tradedisclosure) to the longest horizon (180 days) considered in our analysis. Moreover, the market reaction to the disclosure of trades by bank insiders is stronger than for other firms conditionally on a sell order. Our findings are not driven by times of crisis, when information asymmetries between banks and outside investors might get more severe. If anything, crises appear to have an amplifying effect on our average findings. The difference in the profits on trades of bank and non-bank stocks, in months where insiders are net sellers, becomes even lower during a crisis. To conclude, our evidence of a greater opacity of banks especially with regards to negative information is of particular interest for policy. Regulators should be concerned about the implications of bank opacity for financial stability, because the market in unable to discipline banks following bad news. Our analysis of the determinants of bank opacity suggests that this problem is more severe for banks with higher levels of loan loss allowances.

References Aboody, D., Lev, B., 2000. Information Asymmetry, R&D, and Insider Gains. Journal of Finance 55(6), 2747-66. Adebambo, B., Brockman, P., Yan, X., 2015. Anticipating the 2007 2008 Financial Crisis: Who Knew What and When Did They Know It? Journal of Financial and Quantitative Analysis 50 (04), 647-669. Collin-Dufresne, P., Vos, Y., 2015. Do Prices Reveal the Presence of Informed Trading? Journal of Finance 70 (4), 1555-1582. Cziraki, P., 2015. Trading by Bank Insiders Before and During the 2007-2008 Financial Crisis. Mimeo. Dewally, M., Shao, Y., 2013. Financial derivatives, opacity, and crash risk: Evidence from large US banks. Journal of Financial Stability 9, 565-577. Fahlenbrach,, R, Stulz, R., 2011. Bank CEO Incentives and the Credit Crisis. Journal of Financial Economics 99, 11-26. Flannery, M.J., Kwan, S.H., Nimalendran, M., 2004. Market evidence on the opaqueness of banking firms assets. Journal of Financial Economics 71, 419 460. Flannery, M.J., Kwan, S.H., Nimalendran, M., 2013. The 2007 2009 financial crisis and bank opaqueness. Journal of Financial Intermediation 22, 55 84. Jeng, L., Metrick, A., Zeckhuser, R., 2003. The profits to insider trading: A performanceevaluation perspective. The Review of Economics and Statistics 85 (2), 453-471. Lakonishok, J., Lee, I., 2001. Are Insider Trades Informative? Review of Financial Studies 14 (1), 79-111. Morgan, D., 2002. Rating banks: risk and uncertainty in an opaque industry. Amererican Economic Review 92, 874 888. Seyhun, N., 1986. Insiders profits, costs of trading, and market efficiency. Journal of Financial Economics 16, 189 212. Seyhun, N., 1992. The effectiveness of the insider-trading sanctions, Journal of Law and Economics 35, 149 182. Wagner, W., 2007. Financial development and the opacity of banks. Economics Letters 97, 6 10.

TABLE 1.A SUMMARY STATISTICS: MONTHLY VARIABLES Non-Banks & Negative Net Purchases Banks & Negative Net Purchases Non-Banks & Positive Net Purchases Banks & Negative Net Purchases VARIABLES mean sd N mean sd N mean sd N mean sd N CAR_TtoDptd -0.0102 0.234 219-0.0473 0.381 199 0.0156 0.231 230 0.0461 0.326 199 CAR_15ptd 0.0300 0.209 219 0.0198 0.236 199 0.00877 0.205 230 0.00821 0.225 199 CAR_30ptd 0.0261 0.186 219 0.0233 0.210 199 0.0134 0.185 230 0.0187 0.196 199 CAR_60ptd 0.0229 0.141 219 0.0184 0.149 199 0.0164 0.139 230 0.0183 0.148 199 CAR_90ptd 0.0238 0.120 219 0.0199 0.129 199 0.0193 0.121 230 0.0201 0.129 199 CAR_180ptd 0.0221 0.0910 219 0.0200 0.0967 199 0.0191 0.0907 230 0.0204 0.0974 199 P_TtoD -80,738 1.497e+06 219 68,429 711,889 199 33,140 290,624 230 6,214 25,938 199 P_15-154,044 1.367e+06 219 65,316 754,867 199 15,414 111,590 230 2,441 23,042 199 P_30-124,266 1.163e+06 219-15,966 373,367 199 12,938 93,342 230 1,807 15,902 199 P_60-118,952 737,455 219-2,516 409,104 199 8,392 42,061 230 1,354 10,951 199 P_90-110,725 550,359 219-27,436 270,351 199 6,050 25,370 230 1,036 9,522 199 P_180-107,192 419,175 219-36,774 230,114 199 7,284 41,550 230 936.8 6,676 199

TABLE 1.B SUMMARY STATISTICS: QUARTERLY VARIABLES VARIABLES mean sd N CAR_TtoD -0.0598 0.733 5,484 CAR_15ptd -0.0508 0.292 5,691 CAR_30ptd -0.0482 0.215 5,792 CAR_60ptd -0.0476 0.154 5,792 CAR_90ptd -0.0450 0.132 5,792 CAR_180ptd -0.0440 0.101 5,792 le 18.25 532.7 5,681 llae 0.276 6.792 5,681 trade 0.241 1.039 498 reale 0.0783 0.309 3,703 opaqe 0.619 16.73 5,789 profe 0.184 10.26 5,681 nin 73.15 37.04 5,683 mlev 2,449 68,420 5,789 Net_P 0.563 0.496 5,792

TABLE 2 BUY-AND-HOLD RETURNS AT DIFFERENT HORIZONS Buy-and-Hold Returns at: Trade-Disclosure 15 Days 30 Days 60 Days 90 Days 180 Days Panel A: Months with Positive Net Purchases mktrf -0.228*** -0.00608-0.0353* -0.00476-0.0152* -0.00451 (0.0832) (0.0331) (0.0201) (0.0131) (0.00844) (0.00672) smb -0.186** 0.0171 0.0299 0.0272 0.0149 0.00923 (0.0830) (0.0541) (0.0250) (0.0169) (0.0144) (0.00915) hml -0.293** -0.0737-0.0550* 0.0125-0.00217 0.00527 (0.116) (0.0544) (0.0298) (0.0205) (0.0148) (0.00917) Constant 0.0396*** -0.00339 0.00414-0.00140-0.000141-0.000450 (0.0141) (0.00653) (0.00342) (0.00213) (0.00146) (0.000987) Observations 195 195 195 195 195 195 R-squared 0.115 0.021 0.072 0.024 0.030 0.012 Panel B: Months with Negative Net Purchases mktrf 0.362** 0.114*** -0.000852 0.0170 0.00497 0.000112 (0.179) (0.0426) (0.0346) (0.0205) (0.0139) (0.0140) smb -0.0805 0.0321 0.0317 0.0339** 0.0116 0.0193 (0.124) (0.0463) (0.0342) (0.0156) (0.0147) (0.0125) hml 0.117 0.0578-0.0145 0.0273 0.0231 0.0227 (0.219) (0.0637) (0.0366) (0.0315) (0.0183) (0.0144) Constant -0.0512* -0.0150* -0.00348-0.00612** -0.00576*** -0.00330 (0.0264) (0.00841) (0.00465) (0.00310) (0.00221) (0.00208) Observations 197 197 197 197 197 197 R-squared 0.058 0.059 0.011 0.037 0.015 0.022

TABLE 3 PROFITS AT DIFFERENT HORIZONS Profits at: Trade- Disclosure 15 Days 30 Days 60 Days 90 Days 180 Days Panel A: Months with Positive Net Purchases mktrf 101,112* 52,698** 33,212* 18,492 8,101 13,825 (53,890) (26,544) (19,566) (12,921) (6,101) (9,192) smb -41,436-25,719-6,372-3,610-7,013-18,065 (41,928) (20,781) (23,997) (16,213) (9,600) (15,744) hml -190,425-56,639-71,840-26,920* -11,191-7,652 (181,822) (68,009) (54,144) (15,460) (7,396) (10,442) Constant -34,489-16,473* -13,694* -8,954*** -6,080*** -7,851** (22,473) (8,766) (7,120) (3,157) (1,689) (3,063) Observations 195 195 195 195 195 195 R-squared 0.014 0.014 0.019 0.020 0.011 0.008 Panel B: Months with Negative Net Purchases mktrf -2.856e+06*** - 2.492e+06*** - 2.026e+06*** - 1.011e+06** - 513,045*** -182,605 (906,413) (813,215) (757,513) (435,140) (181,813) (144,203) smb -1.522e+06** -526,832-956,870* -262,598 68,527 140,547 (725,544) (758,559) (538,350) (455,372) (406,777) (199,320) hml 246,448 634,518 1.364e+06*** 444,151 368,127 207,366 (692,872) (656,331) (473,477) (330,431) (245,963) (202,073) Constant 255,020** 313,992*** 176,717** 159,776*** 109,465*** 86,083*** (115,198) (102,069) (84,011) (53,927) (32,884) (25,649) Observations 197 197 197 197 197 197 R-squared 0.211 0.168 0.275 0.129 0.078 0.021

TABLE 4 BUY-AND-HOLD RETURNS AFTER DISCLOSURE Buy-and-hold return at: Disclosure + 1 Day Disclosure + 2 Days Disclosure + 3 Days Panel A: Months with Positive Net Purchases bank 0.0611 0.0508-0.00214 (0.0403) (0.0378) (0.0362) Constant -0.105*** -0.0913*** -0.0430*** (0.0165) (0.0155) (0.0148) Observations 7,093 7,097 7,098 R-squared 0.000 0.000 0.000 Panel B: Months with Negative Net Purchases bank -0.128*** -0.100** -0.0567 (0.0439) (0.0395) (0.0382) Constant 0.00442-0.00440 0.00884** (0.00552) (0.00504) (0.00450) Observations 22,495 22,504 22,504 R-squared 0.001 0.000 0.000

TABLE 5 BUY-AND-HOLD RETURNS AT DIFFERENT HORIZONS: CRISIS VS. NORMAL TIMES Buy-and-Hold Returns at: Trade-Disclosure 15 Days 30 Days 60 Days 90 Days 180 Days Panel A: Months with Positive Net Purchases mktrf -0.230*** -0.00670-0.0310-0.00564-0.0151* -0.00451 (0.0840) (0.0332) (0.0196) (0.0133) (0.00869) (0.00669) smb -0.185** 0.0174 0.0276 0.0277 0.0149 0.00923 (0.0806) (0.0543) (0.0244) (0.0170) (0.0144) (0.00914) hml -0.294** -0.0737-0.0543* 0.0124-0.00216 0.00527 (0.117) (0.0544) (0.0283) (0.0205) (0.0148) (0.00920) crisis -0.00894-0.00277 0.0196** -0.00395 0.000190-1.98e-05 (0.0661) (0.0180) (0.00892) (0.00609) (0.00527) (0.00270) Constant 0.0408*** -0.00301 0.00148-0.000859-0.000167-0.000448 (0.0131) (0.00702) (0.00367) (0.00229) (0.00157) (0.00103) Observations 195 195 195 195 195 195 R-squared 0.115 0.022 0.094 0.026 0.030 0.012 Panel B: Months with Negative Net Purchases mktrf 0.300** 0.108** -0.000513 0.0158 0.00559 7.14e-05 (0.147) (0.0414) (0.0355) (0.0196) (0.0147) (0.0140) smb -0.0465 0.0357 0.0315 0.0346** 0.0113 0.0193 (0.115) (0.0478) (0.0345) (0.0160) (0.0144) (0.0124) hml 0.105 0.0565-0.0144 0.0270 0.0232 0.0227 (0.190) (0.0635) (0.0366) (0.0310) (0.0184) (0.0144) crisis -0.294*** -0.0311 0.00161-0.00584 0.00295-0.000194 (0.102) (0.0259) (0.0164) (0.0132) (0.00824) (0.00549) Constant -0.0119-0.0109-0.00369-0.00534** -0.00615** -0.00328 (0.0222) (0.00876) (0.00497) (0.00265) (0.00238) (0.00220) Observations 197 197 197 197 197 197 R-squared 0.152 0.068 0.011 0.040 0.016 0.022

TABLE 6 PROFITS AT DIFFERENT HORIZONS: CRISIS VS. NORMAL TIMES Profits at: Trade-Disclosure 15 Days 30 Days 60 Days 90 Days 180 Days Panel A: Months with Positive Net Purchases mktrf 96,077 48,620* 28,180 12,240 4,133 5,764 (58,476) (27,742) (21,202) (12,664) (6,792) (10,512) smb -38,784-23,571-3,722-317.3-4,923-13,819 (42,642) (20,832) (23,373) (14,819) (8,641) (13,308) hml -191,214-57,278-72,628-27,900-11,813-8,915 (181,640) (68,212) (54,620) (17,153) (8,446) (14,066) crisis -22,740-18,422-22,726-28,237-17,922* -36,407* (43,678) (20,527) (22,868) (18,383) (9,714) (21,655) Constant -31,398-13,969-10,605-5,116** -3,644*** -2,903** (26,018) (9,934) (7,765) (2,257) (1,300) (1,305) Observations 195 195 195 195 195 195 R-squared 0.015 0.017 0.025 0.068 0.073 0.089 Panel B: Months with Negative Net Purchases mktrf -2.873e+06*** -2.489e+06*** -1.991e+06** -961,070** -461,884** -122,789 (918,118) (824,549) (781,838) (448,049) (187,835) (147,861) smb -1.513e+06** -528,543-976,090* -289,862 40,417 107,681 (725,185) (758,659) (533,443) (462,879) (418,457) (206,552) hml 243,098 635,160 1.371e+06*** 454,380 378,673 219,697 (692,807) (657,520) (476,095) (344,433) (261,715) (223,655) crisis -77,367 14,824 166,484 236,163 243,493* 284,686*** (311,473) (215,025) (206,473) (157,593) (130,335) (108,948) Constant 265,373** 312,009*** 154,440 128,176** 76,884** 47,990** (126,594) (114,298) (95,765) (58,149) (31,407) (23,296) Observations 197 197 197 197 197 197 R-squared 0.211 0.168 0.277 0.140 0.108 0.091

TABLE 7 BANK LEVEL DETERMINANTS OF INSIDER GAINS Panel A: Months with Positive Net Purchases Panel B: Months with Negative Net Purchases Buy-and-Hold Returns at: Trade-Disclosure 15 Days 30 Days 60 Days 90 Days 180 Days Trade-Disclosure 15 Days 30 Days 60 Days 90 Days 180 Days le 0.197 0.0366 0.0338 0.0568*** 0.0549*** 0.0690*** -0.233 0.276 0.0897 0.0891 0.158 0.0873 (0.142) (0.0490) (0.0324) (0.0198) (0.0186) (0.0137) (0.741) (0.275) (0.181) (0.133) (0.118) (0.0850) llae -0.532-0.472-0.104-0.270-0.178-0.533** 10.95 2.771-4.932* -4.724*** -2.688** -2.706** (1.399) (0.375) (0.289) (0.195) (0.266) (0.232) (13.03) (4.240) (2.733) (1.607) (1.221) (1.133) trade 0.0639 0.0439 0.0698 0.0976* 0.0898* 0.125*** -1.032 0.348 0.0597 0.147 0.202 0.260** (0.294) (0.118) (0.0918) (0.0534) (0.0507) (0.0376) (1.150) (0.395) (0.267) (0.195) (0.177) (0.122) reale 3.089** 1.027 0.159 0.124 0.0306-0.107-30.18-8.552-3.164-3.321-5.671** -4.729** (1.452) (0.641) (0.491) (0.334) (0.309) (0.235) (18.45) (5.246) (3.685) (2.344) (2.292) (1.944) opaqe 1.093 0.115 0.0739-0.128-0.134-0.159-4.485** 0.754 0.305 0.178 0.245 0.0618 (0.955) (0.327) (0.282) (0.154) (0.158) (0.108) (2.171) (0.887) (0.440) (0.299) (0.255) (0.195) profe 1.058*** -0.176 0.0670-0.00387-0.0802-0.00159 3.513** 0.342-0.206-0.162-0.130-0.0988 (0.395) (0.124) (0.0971) (0.0614) (0.0650) (0.0530) (1.545) (0.580) (0.273) (0.199) (0.205) (0.126) nin 0.000908 0.000732 0.00155*** 0.00107*** 0.000611* 0.00159*** -0.00708 0.000864 0.000996 0.00104 0.000706 0.000896* (0.00231) (0.000754) (0.000506) (0.000404) (0.000355) (0.000291) (0.00533) (0.00176) (0.00131) (0.000916) (0.000742) (0.000511) mlev -0.00212** -0.000183-0.000283-0.000362** -0.000364** -0.000386*** 0.00627-0.00138 0.000867 0.000174-0.000776-0.000337 (0.00102) (0.000390) (0.000268) (0.000155) (0.000141) (9.53e-05) (0.00693) (0.00230) (0.00141) (0.00105) (0.000957) (0.000675) pinv1 6.040-1.097-0.0727-0.492-0.685-1.180*** 3.280-4.143-5.928 4.473 6.161* 3.506 (4.148) (1.342) (0.741) (0.508) (0.479) (0.387) (29.89) (13.61) (5.504) (3.950) (3.263) (2.824) LMV1 0.217 0.205 0.0849 0.107* 0.0198 0.0701 0.311 0.803 0.253 0.0715-0.00799-0.0851 (0.326) (0.132) (0.0962) (0.0647) (0.0600) (0.0496) (1.635) (0.737) (0.386) (0.285) (0.211) (0.150) Constant -2.851-2.653-1.127-1.381-0.184-0.883-7.140-12.16-4.301-1.545-0.128 1.099 (4.373) (1.762) (1.279) (0.864) (0.806) (0.663) (25.54) (11.35) (5.962) (4.392) (3.245) (2.327) Observations 303 313 333 333 333 333 153 165 165 165 165 165 R-squared 0.392 0.436 0.378 0.379 0.331 0.447 0.418 0.401 0.386 0.404 0.385 0.431