Political engagement
and the bailout of financial firms
Florian Peters*
1Master Thesis - MSc in Finance
University of Groningen
Supervisor: Dr. M. Hernandez Tinoco**
2Abstract
Financial firms which are active in political engagements, such as lobbying or personal
connections to politics, seek preferential regulatory treatment during periods of economic
crises. This study investigates the relation between politically engaged firms and the
likelihood, timing and level of receiving bailout money under the Troubled Asset Relief
Program in 2008 (TARP). The results reveal that politically-engaged financial firms have a
higher chance of receiving TARP support and to receive a larger amount of TARP support.
Firms that both lobbied and maintained a personal political connection prior to TARP,
received bailout funds more timely.
Key words: Political Engagement, Lobbying, Political Connection, Bailout
*1 Student at the Faculty of Economics and Business, University of Groningen, The Netherlands E-Mail: florian_peters@web.de, student number S2574675
2
1. Introduction
“In most of the cases public interests are usually subject to private interests” (Sallust3, 86 – 36 BC). What has already been noticed 2,000 years before is more topical than ever in the sense of the latest financial crisis of 2007-2008 and a fundamental part of research on political economy: The relation between political connections and private economic benefits. Mainly, this topic is relevant in terms of fair market conditions, market efficiency and altruistic governmental decisions. This paper tries to identify market frictions in one of the most market based countries in the world by investigating whether politically engaged firms in the U.S. experience preferential regulatory treatment.
Recent research shows that politically engaged firms are rent seeker and strive to obtain certain economic advantages as a counterperformance (Roberts, 1990; Johnson and Mitton, 2003; Cull and Xu, 2005; Fisman, 2001; Faccio, 2006; Jayachandran, 2006; Gul, 2006; Faccio and Parsley, 2009; Goldman et al., 2009; Cooper et al., 2010; Veronesi and Zingales, 2010; Aggarwal et al., 2011; Duchin and Sosyura, 2012; Bliss and Gul, 2012a,b; Hill et al., 2013).Furthermore, another trend within the literature reveals that firms which are active in political engagements, such as lobbying or personal connections to politics, receive additional advantages during periods of economic distress. The latest study (Blau et al., 2013) regarded to this topic is based on Faccio et al. (2006) and shows that politically engaged firms are more likely to receive bailout funds than non-engaged peers. This study extends the work of Blau et al. (2013) by testing the relation between political engaged firms and the probability of receiving bailout money in comparison to non-engaged firms with a larger hand-collected data set and adding political campaign contributions as a magnitude to proxy lobbying. With regard to the data this study includes 84 more firms than the data base of Blau et al. (2013) which encompass 97% of the total bailout funds distributed by the U.S. Treasury Department from 2008 to 2009, whereas Blau et al. (2013) take 79% into account. In addition, all firms have been checked for subsidiaries, mergers, acquisitions or name changes that give indices for lobbying activities or direct personal political connections. And, to obtain all lobbying activities political campaign contributions are taken into account which adds on average $573,732 to the lobbying expenditure of each firm that lobby (on average $6.2 million). Hence, in total 21 more lobbying firms are identified in comparison to Blau et al. (2013).
The focus rests on the Troubled Asset Relief Program (TARP) of 2008 and its allocation of bailout money. Furthermore, this study investigates whether the amount of TARP-money received by politically engaged firms is larger or distributed more timely.
Since the distribution process of TARP bailouts by the U.S. Treasury Department is unknown it is even more valuable to research any correlation between political connection and TARP bailouts.
3
3
The only criteria mentioned by “The New York Times” is the CAMELS4 rating system. This rating classifies approximately 8,500 U.S. banks into five categories in which a ranking of 1 applies to those with the best economic conditions and a ranking of 5 to those with the worst economic condition5. In the manner of U.S. Treasury Henry M. Paulson the program is preferred for financial institution with a sound economic basis. Besides political engaged variables, several control variables are additionally included in the model to ensure a thorough examination of which firm characteristics might have an impact on the likelihood of receiving TARP funds. Hence, firm characteristics like market cap, total assets, leverage ratio, share turnover or idiosyncratic firm risk might have explanatory power as well. However, the main focus remains on political engagement and the relation within the distribution process of TARP money provided by the U.S. Treasury Department. Therefore, three hypotheses are tested:
(1) Political engagement of financial firms leads to a higher probability of receiving TARP support.
(2) Political engagement of financial firms leads to a more timely payout of TARP money. (3) Political engagement of financial firms has an influence on the level of TARP support.
Political engagement is classified and measured in two realms. First, direct money contribution are considered, which are proxied with lobbying donations and campaign contributions6 (Cooper et al., 2010; Yu and Yu, 2011). Second, following Faccio et al. (2006), political engagement is proxied with the number of political connections of a financial firm (Blau et al., 2013). According to the Center for Responsive Politics (CRP), a firm is politically connected if at least one person has an employment history in the federal government and afterwards in the firm or vice versa or a concurrently combination of both7. Therefore, political engagement serves as a generic term to describe firms that lobby and have political connections, whereas lobbying includes direct political donations and indirect campaign contributions (PACs).
The results reported in this study are in line with those of Blau et al. (2013). First, statistically significant results are found that lobbying firms have a higher probability of receiving TARP funds than non-lobbying firms. The marginal impact on receiving TARP money averages 15% for lobbying firms at the 0.1 level. On the contrary, Blau et al. (2013) find a statistically significant marginal impact of 42% at a 0.01 level. Furthermore, the results show that firms that receive TARP funds have a financial size (market cap and total assets) that is 3 times larger than non-TARP firms. In addition the marginal impact of financial size is approximately 10% and significant at the 0.01 level.
4
CAMELS rating system: Capital adequacy, Assets, Management Capability, Earnings, Liquidity, Sensitivity
5http://www.nytimes.com/2008/11/01/business/01bank.html 6
PACs – Political Action Committees
7
4
However, on average the financial size of lobbying firms that receive TARP support is more than 13 times larger than non-lobbying firms within the TARP-sample. Therefore, it seems that financially large firms are more prone to lobby than smaller firms. Further, larger firms are more systemic relevant and therefore do more likely receive TARP funds. The question if financial size leads to a higher probability of investing money in political connections or vice versa might be an interesting topic to investigate further in future research. In other words: Does financial size lead to political contributions or political contribution to financial size? In addition, firms that are politically connected according to CRP had a 31% higher probability of receiving bailout money than non-connected firms (p-value = 0.000). This result is supported by the fact that the number of political connected firms which receive bailout money is approximately 3 times larger than connected firms that do not receive bailout support. This is in line with Blau et al. (2013), who find a marginal impact of 29%.
The next tests are aimed to find out whether political engaged firms do receive a more timely payout of TARP money. Results of the summary statistics reveal that TARP payout is distributed over 34 days. On the beginning date on October 28th, 2008, eight firms received a total amount of $125 billion (63% of total TARP payout).Within the first two payout dates half of all lobbying and political connected firms are presented. Also within the first ten payout dates 91% of all lobbying and 92% of all political connected firms received 96% of total bailout money. To test whether this observation is statistically valid, a negative binomial regression is used. The results reveal that firms which both lobby and maintain political connections do receive TARP money 41% more timely than other firms. No statistically significant results are found for firms that either lobby or maintain political connections. Again, financial size seems to matter at a significance level of 0.01. The results suggest that firms with higher financial size received TARP money 13% to 16% more timely than firms with lower market capitalization or total assets. The second hypothesis stated in the hypothesis is therefore only partly confirmed due to the fact that only firms that concurrently lobby and maintain political connections receive TARP money more timely and that financial size bias the results.
5
Firms receive on average $396.32 billion for every dollar spent on lobbying during the 5 years prior to the beginning of TARP in 2008. However, the results are not solely driven by lobbying expenditures. Political connections do have an influence on the marginal effects of lobbying donations and TARP support as well (p-value 0.000).
The remaining structure of the paper includes a holistic literature review in section 2 and a description of the TARP and the decision why political campaign contributions are added as a lobbying factor. Section 3 comprises a description of the sample construction and explanations of the main variables. In Section 4 the empirical results are analyzed and concluded in section 5.
2. Literature review
The literature reveals three main trends within the scope of political engagement of firms. First, Stigler (1971), and Peltzman (1976, 1989) investigate the interaction between corporate political connections and beneficial government legislation. They elaborated a theory of economic regulation aiming to explain which firm will likely receive governmental benefits during a period of regulation and to investigate the general magnitude of impact of the state on firms. Due to the deregulation period since 1980 the theory of regulation has not been investigated further. In the same context De Soto (1989) and Backman (1999) demonstrate a significant relation between political engagement and preferential tax regimes, and preferential treatment by government owned enterprises.
Second, a wide range of studies provide insights into the relation between political engagement and firm performance8. Faccio (2006) examines firms in 47 countries and investigates the link between connected firms and their value measured by stock price. The findings reveal a positive effect of politically corporate connections on the firm`s stock price, which decrease when the government has set more limits on official behavior. These results are in line with Fisman (2001) who estimate the value of political engaged firms in Indonesia. Jayachandran (2006) studies a comprehensive impact of political connected firms on their stock market value. His event study is based on the surprising withdrawal of Senator James Jeffords from the Republican Party in May 2001.
The results show that a firm lost 0.8 percent of market capitalization, the week of Jeffords’s switch to the Democratic Party, for every $250.000 it gave to the Republicans in the previous election cycle. On the other hand firms which contributed to the Democratic Party faced an increase in their market value. In contrast Gul (2006) finds adverse effects of political connected firms in times of political crisis.
8Additional literature Roberts (1990), Cull and Xu (2005), Faccio and Parsley (2009), Goldman et al. (2009), Cooper et al.
6
His results show that connected firms have greater audit fees than non-connected firms as a consequence of the Asian financial crisis. This circumstance is related to the perception of less transparency and economic fragility of connected firms when Malaysian government could not financially assist those firms directly after the turmoil. However, when capital controls were implemented, audit fees of connected firms declined significantly than other non-connected firms. Similar results are provided by Johnson and Mitton (2003) and Fraser et al. (2006). Bliss and Gul (2012a,b) extended the work and reveal a positive effect between leverage and political nepotism. They show that connected firms in Malaysia have greater leverage and higher costs of debt which lead to inherently higher risks.
7
2.1. Corporate lobbying and political campaign contributions
A main contribution of this paper is the inclusion of campaign contributions as an additional political corporate lobbying facto, for instance, Faccio (2006) proxy political connections of firms when they have registered political action committee (PAC). According to CRP, political action committees (PACs) are defined as a campaign financial system in which the donated money to politicians or political parties originates from individuals like firm directors, employees, or their families to support candidates for elections up to a maximum of $5,000 per candidate per election (Hill et al., 2013). The reason behind the indirect way is that firms are not allowed to make direct contributions to political campaigns from the corporate treasury. In contradiction to PACs, lobbying donations are not limited and can be initiated directly from the corporate treasury (idem.). Since both types of corporate politically engagements indicate contributions to political candidates, parties or even lobbyists it is reasonable to include PACs as a further or indirect magnitude of lobbying. This procedure is confirmed by Hill et al. (2013) who extend the work of Cooper et al. (2010) by using both types of corporate lobbying to show an influence of firm value and corporate lobbying.
2.2. Troubled Asset Relief Program (TARP)
On October 3rd, 2008, the TARP bill was signed into law to address the escalating financial crisis at that time (de jure: Emergency Economic Stabilization Act). TARP, proposed and implemented by the Treasury department, aimed for stabilizing the U.S. financial system by purchasing approximately $700 billion worth of toxic assets (i.e. mortgage backed securities or insolvent bank loans). However, on October 14th, 2008 a revised plan was announced by the government. Revisions to the TARP intended to invest up to $250 billion of the available $700 billion to invest in preferred stocks in U.S. based financial institutions under the Capital Purchase Program (CCP).
8
3. Sample construction and main variables
The sample selection process is subdivided in three steps. Step one includes the identification of all active banks located in the U.S. via bankscope. In total 9,694 banks are identified with a respective ISIN and bankscope identification number. ISIN stands for International Securities Identification Number and indicates a listed bank.
Step two encompasses the data selection process of firms that received TARP funds. The U.S. Treasury Department discloses “Transaction Reports” for each payout date of the investment program TARP, which contains a list of all investments distributed to financial firms. After compiling the data, in total $204.9 billion have been invested in 707 financial firms based in the U.S. Furthermore, the information for the amount repaid by each bank is gathered as well.
Step three comprises the detection of firms that lobbied or maintain any other type of political connection. Following Blau et al. (2013) data on the lobbying donations and campaign contributions of financial firms located in the U.S. is gathered from the Center for Responsive Politics (CRP). The lobby data set consists of yearly lobbying disclosure reports and includes hard, soft, and basic lobbying donations and campaign contributions for each financial firm (Blau et al., 2013). The magnitude of measuring lobbying donations and campaigns contributions is an aggregate measure and does not contain the respective legislators who were lobbied. In this case, an aggregate measure seems to be sufficient since the recipients of TARP exhibit similar contribution patterns (idem., Kroszner and Stratman, 1998). The “Revolving Door” database of CRP provides information whether a financial firm has political connections. As already described in the introduction, a firm is politically connected if at least one person has an employment history in the federal government and afterwards in the firm or vice versa or a concurrently combination of both. Following Blau et al. (2013), lobbying donations and contributions via PACs represent the sum of the 5 years prior to the beginning of TARP in 2008, which proxies the second legislature of the Republicans. This study adds firms that lobbied also in the first legislator of the Republicans beginning in 2001 and investigates whether the results are robust to those reported in section 49.
After following these three steps, the list including all active 9,694 banks is used to match the TARP banks and lobbying banks with the respective ISIN identifier to sort out all listed firms. Furthermore, all firms have been checked for subsidiaries, mergers, acquisitions or name changes that give indices for lobbying activities or direct personal political connections.
9
9
Firm specific data is gathered via Datastream. Non-listed banks and banks which data is not completely available via Datastream or bankscope are omitted. The reason why the focus is set on listed banks is to control for market cap as a second indicator of financial size next to total assets. A further reason is to control for idiosyncratic risk of each firm on the market and the daily shares traded (share turnover). Since the shares traded each day during 2007 is an indicator for the involvement of a financial firm in the subprime mortgages crises. The higher the share turnover during this period the higher the chance that a firm is part in the spiral of trading toxic assets (i.e. assets mortgage backed securities or credit default swaps) and more clearly reveal heightened funding and counterparty risk. The idea is based on the work of Eichengreen at al. (2009) who find a common factor in the trading activity of banks' credit default swap spreads.
As already mentioned in the introduction, these variables control for firm characteristics which may have an impact on the likelihood of receiving TARP funds and the magnitude of TARP funds received. This procedure leads to a TARP-sample of 294 banks and a control group-sample of 361 banks. Within the TARP-sample, 35 banks lobbied or contributed to PACs, whereas 17 banks did within the control group-sample. Furthermore, 52 banks have direct political connections, according to CRP, within the TARP-sample and 27 banks within the control-group-sample.
Panel A in Table 1 shows summary statistics of firms who received TARP money. To compare TARP-firms with TARP-firms that did not receive TARP support, Panel B reports the control-group-sample. The data represent the average figures of the balance sheet over the entire business year prior to the TARP payout date. For example, if the TARP payout date of one firm was in 2008 (2009), the particular figures are related to 2007 (2008). The assumption behind this procedure is that the U.S. Treasury Department especially related to the most recent balance sheet figures of firms while deciding to which firm and how much TARP funds should have been distributed. On average a TARP-firm has a stock price (Price) of $51.01, a market capitalization (Size) of $2.16 billion, assets (TotAssets) of $32.60 billion, and a debt-to-equity (D/E) ratio of 20.97. The amount of TARP funds received averages $676.02 million. Furthermore the average share turnover or the percentage of shares outstanding that are traded daily in 2007 is 0.21%. The idiosyncratic risk or daily share turnover averages 3.23%10.
Following Blau et al. (2013), and transform the data using natural log of the variable Size (skewness = 0.90; kurtosis = 3.96) and TotAssets (skewness = 1.35; kurtosis = 5.40), evidence for non-normality condition is detected.
10
10
After winsorizing the outlier twice, the distribution properties of the variable Size (skewness = 0.75; kurtosis = 3.32) and TotAssets (skewness = 1.26; kurtosis = 5.04) only slightly changed.
PACDUM indicates a firm with one, if positive lobbying contributions are identified during 5 years
prior to their receipt of bailout funds, and zero otherwise. The other discrete variable CONNECT indicates a firm with one if a political connection according to CRP is reported, and zero otherwise. The results show that 11.90% of firms that received TARP funds lobbied and 18.90% maintain direct political connections. The continuous variable PACDol indicates lobbying donations and contributions to PACs of each firm that lobbied. On average, each lobbying firm spent $6.71 million on lobbying during the second legislature (2003 to 2008). Firms that did not receive TARP funds (Panel B), on average have a Price of $62.71, a Size of $0.84 billion, TotAssets of $11.60 billion, a D/E of 20.58, Turn of 0.28%, and Volt of 3.72%. In addition, 4.71% firms lobbied of all firms which did not receive TARP money and 7.48% have direct political connection. Those firms that lobbied contributed on average $0.37 million.
Panel C identifies the differences in means and its statistical significance. It is obvious that financial size appears to have an impact on the likelihood of receiving TARP funds. Both financial size variables (Size and TotAssets) are significantly different at a 0.01 level. Furthermore, TARP-firms have a statistically significant lower share turnover (difference = 0.06%; p-value = 0.036). This result may indicate that banks who anticipated less in speculative trades are more likely to receive TARP funds. A further interesting result is that firms that lobbied (PACDUM), spent more money on lobbying (PACDol) or have political connections (CONNECT), were more likely to receive bailout dollars (at a significance level of 0.01).
11
Table 1
Summary statistics.
The table shows summary statistics of the sample and is subdivided into three Panels. Panel A illustrates all firms that received TARP funds and Panel B shows an overview of the non-TARP sample. The variables chosen according to Blau et al. (2013) are gathered via Datastream. In particular price (Price), the market capitalization (Size) in millions, the total assets in millions (TotAssets), Debt-to-Equity ratio (D/E), the share turnover (Turn), which indicates the percentage of daily traded volume with respect to shares outstanding and the idiosyncratic volatility (Volt), which is the standard deviation of the residuals from the daily CAPM regressions. PACDUM indicates firm i with one if positive lobbying contributions are identified during 5 years prior to their receipt of bailout funds, and zero otherwise. The other discrete variable CONNECT indicates firm i with one if a political connection according to the Center for Responsive Politics (CRP) is reported, and zero otherwise. The continuous variable PACDol indicates lobbying donations and contributions to PACs of each firm that lobbied. Of the 294 firms that received TARP funds, 35 of the firms spend money on lobbying or PACs and 53 firms maintain political connection according to the Center for Responsive Politics (CRP). Of the 364 firms that did not receive TARP funds, only 17 firms spent money on lobbying or PACs and 27 firms are identified having political connections according to CRP.
Price Size
(in mil.)
TotAssets
(in mil.)
D/E Turn Volt TARP
(in mil.)
PACDUM PACDol Connect
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
Panel A. Bailed out firms characteristics (N = 294)
Mean 51.01 2,159.60 32,599.95 20.97 0.2143 0.0323 676.02 0.1190 6,708,250 0.1803 Min 4.33 7.57 67.30 0.00 0.0004 0.0178 1.00 0.0000 500 0.0000 25th 13.12 35.83 560.42 11.47 0.0368 0.0277 11.43 0.0000 7,263 0.0000 Median 19.01 102.55 1,288.86 17.88 0.0933 0.0320 30.00 0.0000 146.450 0.0000 75th 31.94 534.85 4,377.01 24.78 0.3337 0.0345 93.75 0.0000 1.088.791 0.0000 Max 2,549.00 55,878.15 1,715,746.00 139.52 1.7220 0.0657 25,000.00 1.0000 57.975.127 1.0000
Panel B. Non-bailed out firms characteristics (N = 361)
Mean 62.71 837.41 11,620.24 20.58 0.2738 0.0319 - 0.0471 369.436 0.0748 Min 0.82 5.16 12.26 0.00 0.0002 0.0095 - 0.0000 3.150 0.0000 25th 10.49 31.87 469.91 7.48 0.0221 0.0214 - 0.0000 8.500 0.0000 Median 15.93 66.74 906.45 14.13 0.0716 0.0271 - 0.0000 45.750 0.0000 75th 26.37 323.10 2,488.70 23.24 0.3539 0.0384 - 0.0000 89.500 0.0000 Max 4,625.00 27,664.60 848,684.00 487.65 3.2767 0.0922 - 1.0000 5.207.035 1.0000
Panel C. Difference in means
Difference -11.40 1,322.20 20,979.71 0.38 -0.0595 0.0005 0.0720 6,338,815 0.1055
p-Value (0.615) (0.002) (0.030) (0.882) (0.036) (0.611) (0.001) (0.013) (0.000)
12
Table 1.1
Summary statistics.
The table reports statistics that describe the sample of firms that received TARP money. In particular, panel A reports the results for the 35 firms that received bailout money and lobbied while panel B shows the results for financial firms that receive bailout money and did not lobby. The variables chosen according to Blau et al. (2013) are gathered via Datastream. In particular price (Price), the market capitalization (Size) in millions, the total assets in millions (TotAssets), Debt-to-Equity ratio (D/E), the share turnover (Turn), which indicates the percentage of daily traded volume with respect to shares outstanding and the idiosyncratic volatility (Volt), which is the standard deviation of the residuals from the daily CAPM regressions. CONNECT indicates firm i with one if a political connection according to the Center for Responsive Politics (CRP) is reported, and zero otherwise. Of the 35 firms lobbied, 35 of the firms spend money on lobbying or PACs and 53 firms maintain political connection according to the Center for Responsive Politics (CRP).
Price Size
(in mil.)
TotAssets
(in mil.)
D/E Turn Volt TARP
(in mil.)
Connect
[1] [2] [3] [4] [5] [6] [7] [8]
Panel A. Bailed out firms that lobby (N = 35)
Mean 76.79 11,119.09 179,040.71 22.46 0.4330 0.0335 4,108.08 0.7429 Min 4.33 28.43 2,240.09 0.66 0.1372 0.0229 50.00 0.0000 25th 13.12 953.34 7,765.80 12.28 0.3395 0.0299 157.50 0.5000 Median 19.01 2,968.31 37,015.46 19.90 0.3995 0.0323 935.00 1.0000 75th 31.94 16,819.34 135,769.00 27.20 0.5093 0.0367 3,270.82 1.0000 Max 2,549.00 55,878.15 1,715,746.00 73.32 1.0340 0.0448 25,000.00 1.0000
Panel B. Bailed out firms that did not lobby (N = 259)
Mean 47.52 948.86 12,810.66 20.76 0.1848 0.0322 212.23 0.1042 Min 4.33 7.57 67.30 0.00 0.0004 0.0178 1.00 0.0000 25th 12.51 32.94 473.43 11.46 0.0314 0.0275 10.15 0.0000 Median 17.99 80.10 1,005.41 17.16 0.0740 0.0317 24.00 0.0000 75th 27.415 293.23 2,664.97 24.689 0.2403 0.0343 66.00 0.0000 Max 2,549.00 31,366.26 1,045,409.00 139.52 1.7220 0.0657 10,000.00 1.0000
Panel C. Difference in means
Difference 29.27 10,170.23 166,230.05 1.70 0.2482 0.0013 3,895,847,890 0.6386
p-value (0.374) (0.000) (0.000) (0.596) (0.000) (0.310) (0.000) (0.000)
4. Analysis and results
4.1. Political engagement and its level of impact on receiving TARP support
Summary statistics in Table 1 suggests that both lobbying and political connections lead to a higher probability of receiving TARP support. However, it is unclear by what level of impact this probability is driven. Following Blau et al. (2013), a probit regression is used to investigate the marginal impact:
TARPDUMi = β0 + β1Pricei + β2ln(Sizei) + β3ln(TotAssetsi) + β4D/Ei + β5Turni + β6Volti + β7DUMMYi + εi (1)
The sample includes 655 firms in total. The dependent variable separates the sample by assigning one to the 294 firms which received TARP money and zero for those 361 firms who did not. The independent variables include Price, Size, TotAssets, D/E, Turn and Volt, which have been each defined in section 3. The DUMMY variable is defined twofold.
13
Second, DUMMY is determined as CONNECT, which indicates firm i with one if a political connection according to CRP is reported.
Table 2
Probit regression.
The table shows the results to determine the marginal impact of lobbying and -political connected firms by using cross-sectional data. To achieve the marginal magnitude a Probit regression with the following equation was estimated: TARPDUMi = β0 + β1Pricei + β2ln(Sizei) + β3ln(TotAssetsi) + β4D/Ei + β5Turni + β6Volti + β7DUMMYi + εi.The sample includes 655 firms in total. The dependent
variable separates the sample by assigning one to firms which received TARP money and zero for those 361 firms who did not. The independent variables include Price, Size, TotAssets, D/E, Turn and Volt, which have been each defined in section 3. The DUMMY variable is defined twofold. First in column [1] to[4] the DUMMY is determined as PACDUM, which indicates firm i with one if positive lobbying contributions are identified during 5 years prior to their receipt of TRAP funds, and zero otherwise. Second, in column [5] to [8] the DUMMY is determined as CONNECT, which indicates firm i with one if a political connection according to the Center for Responsive Politics, and zero otherwise. Furthermore in unreported tests the variables Size and TotAssets are highly correlated and show significant interaction effects. Hence, a variety of versions to the model are presented to split the correlated variables. By estimating a linear probability model and calculating the variance inflation factors only the combined variables Size and TotAssets indicate 6. In all other scenarios or columns from [2] to [4] and [6] to [8] the variance inflation factors for every variable are well under 2. The standard errors (in brackets) are heteroskedasticity consistent. Significance levels are indicated: * = 10%, ** = 5%, *** = 1%.
Model number [1] [2] [3] [4] [5] [6] [7] [8]
Panel A. Probit regression.
Intercept -4.0650*** (0.708) -0.1880 (0.143) -2.7997*** (0.527) -3.9921*** (0.721) -3.6334*** (0.699) -0.1758 (0.145) -2.4113*** (0.534) -3.5186*** (0.698) Price -0.0002 (0.000) -0.0001 (0.000) -0.0003* (0.000) -0.0002 (0.000) ln(Size) 0.7401 (0.074) 0.2160*** (0.041) 0.0387 (0.070) 0.1935*** (0.042) Ln(TotAssets) 0.2228*** (0.085) 0.2825*** (0.052) 0.2221*** (0.083) 0.2465*** (0.050) D/E 0.0003 (0.002) 0.0011 (0.001) 0.0021 (0.002) -0.0004 (0.002) 0.0001 (0.002) 0.0011 (0.002) 0.0020 (0.002) -0.0003 (0.002) Turn -1.4842*** (0.241) -0.4465*** (0.144) -1.2792*** (0.223) -1.4073*** (0.234) -1.6407*** (0.248) -0.8418*** (0.171) -1.4451*** (0.230) -1.5741*** (0.240) Volt 7.2870 (4.749) 2.8358 (4.023) 11.3198** (4.597) 5.0700 (4.230) 6.4266 (4.787) 3.2436 (4.075) 10.4660*** (4.606) 5.4264 (4.235) PACDUM 0.2444 (0.212) 0.7509*** (0.143) 0.3591* (0.205) 0.2414 (0.214) Connect 0.7571*** (0.207) 1.0879*** (0.181) 0.7626*** (0.201) 0.7409*** (0.204)
Panel B. Marginal impact.
DUMMY [0.0902] [0.2895]*** [0.1334]* [0.0894] [0.2750]*** [0.4087]*** [0.2803]*** [0.2703]*** Price [-0.0001] [-0.0001] [-0.0001]* [-0.0001] Size [0.0273] [0.0805]*** [0.0141] [0.0674]*** TotAssets [0.0822]*** [0.1047]*** [0.0868]*** [0.0899]*** D/E [0.0001] [0.0004] [0.0008] [-0.0002] [0.0002] [0.0004] [0.0007] [-0.0001] Turn [-0.5475]*** [-0.1721]*** [-0.4770]*** [-0.5214]*** [-0.5960]*** [-0.3163]*** [-0.5311]*** [-0.5742]*** Volt [2.6881] [1.0933] [4.2205]** [1.8782] [2.3346] [1.2187] [3.8465]*** [1.9796] LR 55.58 21.58 50.50 50.90 61.04 42.59 56.08 57.08 Pseudo R² 0.0625 0.0234 0.0526 0.0590 0.0761 0.0463 0.0655 0.0726
Table 2 shows the results of the probit regression. The results reveal evidence of heteroscedasticity. Therefore, the standard errors are corrected using White’s correction. Furthermore, in unreported tests the variables Size and TotAssets are highly correlated (ρ~-0.80) and show significant interaction effects, which gives a hint for multicollinearity. Hence, a variety of versions to the model are presented to split the correlated variables.
14
Due to high correlation and relative high variance inflation factors of Size and TotAssets the validity of the results specified in columns [1] and [5] is impaired. But by omitting the variable Size11, the results reported in columns [1] and [5] are similar. In addition, a logit regression model with the same variables included has been estimated. The unreported results are identical.
The estimate of PACDUM for the full model in Column [1] is positive (0.2444) but not significant (p-value = 0.249). The marginal impact of lobbying on the likelihood of receiving TARP support is moderate with 9.02 percent. It seems that only total assets (estimate = 0.2228, p-value = 0.009) and the share turnover (estimate = -1.4842, p-value = 0.000) have an influence on the dependent variable. However, the marginal impact of total assets is 8.22 percent and -54.75 percent for share turnover. In column [2] after controlling for Price, D/E, Turn and Volt, the estimate of PACDUM is 0.7509 and statistically significant at a level of 0.01. The probability of receiving TARP is 0.3836 if a firm had lobbied.
With regard to the distribution of the variable Price in columns [2] and [6] normality condition is violated (skewness = 12.67; kurtosis = 182.05). But if the variable Price in the model in column [2] is omitted the results are identical to those reported. Column [3] controls for the natural log of market capitalization (ln(Size)) instead of Price. The estimate for PACDUM is again positive and significant (estimate = 0.3591, p-value = 0.080). The marginal impact is 13.34 percent. The results in column [4] after controlling for the natural log of total assets (ln(TotAssets)) do not support an influence of lobbying firms on receiving TARP money (estimate = 0.2414, p-value = 0.258). Again it appears that total assets (estimate = 0.2825, p-value = 0.000) and share turnover (estimate = -1.4073, p-value = 0.000) have a higher influence than lobbying. However, the marginal impact of total assets is only 10.47 percent and -52.14 percent for share turnover. Since share turnover has a negative influence on the likelihood of receiving TARP funds it might undergird the argument that economically sound banks are preferred within the allocation process of TARP money.
A holistic consideration of all results found in columns [1] to [4] lead to the inference that lobbying firms indeed have a higher probability of receiving TARP funds than firms that do not lobby.
This point is more convincing while excluding financial size factors (Size and TotAssets) from the model like in column [2]. Due to the average correlation of ρ~-0.44 between financial size and
PACDUM the significance of PACDUM might be influenced. With regard to financial size, the higher
the Size/ TotAssets of a firm the higher the probability of receiving TARP support.
11
15
Nevertheless, these variables have only an impact of 8 to 10 percent. These findings support the first hint in Table 1 in which financial size of TARP-firms is nearly three times larger (Panel A in Table 1) than the size of non-TARP-firms (Panel B in Table 1). As mentioned in the introduction, this issue addresses the question if financially large firms are more prone to lobby than smaller firms. And, larger firms are more systemic relevant and therefore do more likely receive TARP funds. Faccio et al. (2006) find similar differences between lobbying firms and non-lobbying firms in terms of financial size, but do not investigate this point further. Hence, in relation with the results in Table 1.1, future research might address the question if financial size leads to a higher probability of investing money in political connections or vice versa. In other words: Does financial size lead to political contributions or political contribution to financial size? In addition, the results in columns [1] to [4] are in line with the findings of Blau et al. (2013). The authors find that lobbying firms have a 42% higher probability of receiving TARP funds than firms that did not lobby. The results in this study show on average a 15% higher probability of receiving bailout funds for firms that lobby.
In columns [5] to [8], the DUMMY variable indicates the political connection according to CRP (variable Connect, which is defined in section 3). For brevity only the full model in column [5] is discussed. Again, firms with higher total assets were more likely to receive TARP money by 8 to 9 percent. The variable CONNECT is positive and highly significant at the 0.01 level and detect a marginal impact of 27.50 percentage points. In columns [6] to [8] similar results are identified. The results dedicated to the marginal impact of political connection on the likelihood of receiving TARP support are in line with Duchin and Sosyura (2012) and Blau et al. (2013). While Blau et al. (2013) show a marginal impact of 29% for political connected firms, this study reveals an average marginal impact of 31%.
The data for each variable gathered via Datastream have been dedicated to the business year 2007 or 2008. The idea is based on the fact that the U.S. Treasury had a specific look at the most recent figures to examine the economically situation of each financial firm and adequately allocate TARP support. Since the distribution process of TARP money is unknown, a further probit regression was estimated to corroborate the previous results.
For this purpose, the dataset for each variable has been extended by taking the average for each business year from 2003 to 2007 to proxy the past economic process for each firm.
16
4.2. Political engagement and timing
After investigating the impact of political engaged firms on the likelihood of receiving TARP support this subsection examines the timing of those firms within the TARP sample (Panel A in table 1 in section 3). The hypothesis consists of the assumption that lobbying firms or political connected firms according to CRP receive TARP support more timely than non-engaged firms. Table 3 indicates a chronologically order of the bailout date in relation to quantity of banks which receive bailout funds, the sum of the total payout on each day and the quantity of lobbying and -connected firms. Column [1] reveals that TARP payout is distributed over 34 days. On the beginning date on October 28th, 2008, eight firms received a total amount of $125 billion (63% of total TARP payout).Within the first two payout dates half of all lobbying and political connected firms are presented. Further, within the first ten payout dates 91% of all lobbying and 92% of all political connected firms received 96% of total bailout money. It seems that political engagement contributes to a more timely payout of TARP by the U.S. Treasury. To test this hypothesis and to investigate other factors that might influence the TARP payout timing following equation according to Blau et al. (2013) has been estimated:
TimeToTARPi = β0 + β1Pricei + β2ln(Sizei) + β3ln(TotAssetsi) + β4D/Ei + β5Turni + β6Volti + β7PACDUMi +
β7Connecti + β7PACDUM x Connecti + εi (2)
The dependent variable TimeToTARP indicates a discrete count variable of the days after the signing date of October 3rd, 2008 (column [2] in Table 3). The independent variables include Price, ln(Size),
ln(TotAssets), D/E, Turn, and Volt, which have been defined in section 3. The dummy variable PACDUM indicates firm i with one if positive lobbying contributions are identified during 5 years prior
to their receipt of bailout funds, and zero otherwise. The variable CONNECT denotes firm i with one if a political connection according to CRP is reported and PACDUM x CONNECT determines the interaction of PACDUM and CONNECT.
Since the dependent variable TimeToTARP is discrete, an adequate count regression has to be found. A first approach leads to the Poisson regression. However, a critical criterion for a Poisson regression is a distribution in which the mean is equivalent to its variance. But the mean of TimeToTARP is 107.06 and its variance 4,432.90 which violates the critical criterion of a Poisson regression.
17
The variance inflation factor of the remaining variables is well below 2. This problem is mitigated by various scenarios of the regression model (columns [2] to [4]). By excluding either Size or TotAssets, each variable reveals a variance inflation factor of well below 2 as well. Therefore, following Faccio et al. (2006) the variable size is omitted of the full model. Furthermore, the negative binomial
regression is estimated with Huber-White robust estimates of the standard errors to show a more robust model which meet assumptions concerning normality and homoscedasticity.
Table 3
Time series of bailout money.
The table shows the dates of the bailout payout, the days after the signing date of October 3rd, 2008, the number of firms that received bailout money, the sum of the bailout amount that was paid out that day, the number of firms that lobbied during the last five years prior to 2008, and the number of firms that had political connections according to the Center for Responsive Politics. Bailout date Time to
18
Table 4
Negative binomial regressions.
The table shows the results to determine the influence of politically engaged firms on the timing of receiving TARP money by using cross-sectional data of all 294 firms that received TARP funds. To test this influence, a negative binomial regression according to Blau et al. (2013) has been estimated: TARPDUMi = β0 + β1Pricei + β2ln(Sizei) + β3ln(TotAssetsi) + β4D/Ei + β5Turni + β6Volti + β7DUMMYi + εi.The dependent variable TimeToTARP indicates a discrete count variable of the days after the signing date of
October 3rd, 2008 (column [2] in Table 3). The independent variables include Price, ln(Size), ln(TotAssets), D/E, Turn, and Volt, which have been defined in section 3. The dummy variable PACDUM indicates firm i with one if positive lobbying contributions are identified during 5 years prior to their receipt of bailout funds, and zero otherwise. The variable CONNECT denotes firm i with one if a political connection according to CRP is reported and PACDUM x CONNECT determines the interaction of PACDUM and
CONNECT. Furthermore in unreported tests the variables Size and TotAssets are highly correlated (ρ~-0.95) and show significant
interaction effects. Hence, a variety of versions to the model are presented to split the correlated variables. The standard errors (in brackets) are heteroskedasticity consistent. Significance levels are indicated: * = 10%, ** = 5%, *** = 1%.
19
Table 4 shows results of three panels. For brevity the focus lies on the full model in column [1]. Panel A reports the results including the variable PACDUM. The results identify a negative estimate (-0.0833) which lead to the inference that lobbying firms receive TARP support more timely than non-lobbying firms. However, beside column [2] no estimate in each column shows statistically significant results. In column [3], while controlling for Size, the estimate is not even negative (0.0198).
As already identified in Table 2 and described in section 4.1 it seems that financial size of firms, next to political engagement factors, has a high explanatory power regarding the likelihood of receiving TARP funds. In columns [1] and [4] the estimate for TotAssets is negative (-0.1383; -0.1398) and significant at 0.01 level. Column [3] controls for Size and shows similar results (estimate = -0.1755; p-value = 0.000). In addition, the stock return volatility (Volt) has a positive estimate (13.3560) and is statistically significant at the 0.01 level. This observation is similar in the remaining columns [2] to [4] and suggests that a highly volatile firm receives TARP support later than peers. This might be coherent with the statement of the U.S. Treasury Henry M. Paulson that economic sound banks are preferred. Again, the distribution process of the U.S. Treasury Authority of TARP money is unknown. As mentioned in the introduction, the U.S. Treasury Department might have used the CAMELS12 rating. Therefore, future research may address this problem and include variables which replicate the CAMELS rating system, according to Curry et al. (1999), to investigate the drivers of the TARP distribution process.
Almost identical results are found in Panel B while including the variable CONNECT which are not highlighted for the sake of brevity.
Panel C reports the results of the interaction between PACDUM and CONNECT to determine whether the combination of both variables has any influence on the payout timing. In the full model in column [1] the interaction effect has a negative estimate (-0.5131) and is statistically significant at the 0.1 level. Similar results are found in the remaining columns [2] to [4]. These results lead to the inference that firms who both lobby and have direct political connections according to CRP have a higher probability of receiving TARP support more timely. As in the previous subsection, the regression based on equation (2) is conducted while including a broader range of data. This means the scenario is included that the U.S. Treasury Department examined the balance sheet/ economic process of each firm several business years prior to TARP to make an appropriate choice regarding the distribution of TARP funds. The results reported in Table 8 in the appendix are similar to those identified in Table 4.
12
20
Again, the financial size and the volatility seem to matter as well as described before. Furthermore, due to the average13 correlation of ρ~-0.76 between financial size and PACDUM the significance of
PACDUM might be influenced. Between financial size and the variable CONNECT a correlation of ρ~-0.36 is detected.
Considering the correlation and the fact that the statistical significance of the interacted variables
PACDUM and CONNECT is weak at the 0.10 level and strong for financial size (0.01 level) a further
robustness test is conducted. In 2008 the Financial Stability Board published a list of banks that are considered “too-big-to-fail”. Table 9 (appendix) reports the results of the negative binomial regression while excluding those systemic relevant banks of the sample. The results are similar in all Panels beside the fact that the results in column [2] neither for PACDUM nor for CONNECT are statistically significant. In addition, the interaction variable PACDUM x CONNECT is statistically significant at a 0.10 level instead at a 0.05 level in Table 8.
To uncover the economic significance of the financial size (Size and TotAssets), the estimates of the negative binomial regression have to be transformed into percentage differences. Therefore, the expression 100 x (exp βj - 1), where βj is either β2, β3 or β6. For brevity the averages of all three Panels
and columns have been calculated. The percentage difference of Size is -16.37 and -12.93 for
TotAssets, indicating that firms with higher financial size received TARP money 13% to 16% more
timely than firms with lower market capitalization or total assets. The interaction variable PACDUM x
CONNECT indicate an average percentage difference of -41.05. The estimate of the stock return
volatility (Volt) is not transformed into percentage differences due to an unrealistic high estimate.
Taking all results described in this subsection into account, it is inferred that firms which both lobby and maintain political connections do receive TARP money 41% more timely than other firms. No statistically significant results are found for firms that either lobby or maintain political connections, after conducting a robustness test. Again, the firm characteristic financial size contributes to a more timely payout of TARP money as well. The second hypothesis stated in the hypothesis is therefore only partly confirmed due to the fact that only highly political engaged firms receive TARP money more timely and that financial size bias the results. On the contrary Blau et al. (2013) find significant results for lobbying firms and political connected firms but not for the interaction of both variables. The reason for this contradiction might be attributable to a different data basis.
13
21
4.3. Political engagement and the impact on the magnitude of TARP support
After examining the relation of political engagement and the level of impact on receiving TARP support and timing this subsection focuses on how the level of TARP support is influenced by lobbying contributions or political connection according to CRP. As shown in Table 1.1 in section 3, within the sample of the 294 firms who received TARP support political engaged firms get a higher amount of TARP money. To test this hypothesis, the following equation according to Blau et al. (2013) has been estimated:
TARPmoneyi = β0 + β1Pricei + β2ln(Sizei) + β3ln(TotAssetsi) + β4D/Ei + β5Turni + β6Volti + β7PACDUMi +
β7Connecti + β7PACDUM x Connecti + εi (3)
The dependent variable TARPmoney indicates the amount of dollars (in millions) received by firm i as part of TARP. The independent variables include Price, ln(Size), ln(TotAssets), D/E, Turn, and Volt, which we have been defined in section 3. The dummy variable PACDUM indicates firm i with one if positive lobbying contributions are identified during 5 years prior to their receipt of bailout funds, and zero otherwise. The variable CONNECT denotes firm i with one if a political connection according to CRP is reported and PACDUM x CONNECT determines the interaction of PACDUM and CONNECT. The multivariate OLS regression results using cross-sectional data are presented in Table 5. The results show evidence of heteroscedasticity and are therefore corrected with White’s (1980) correction, even though similar results are identified without the correction (unreported). Furthermore, in unreported tests the variables Size and TotAssets are highly correlated (ρ~-0.72) and show significant interaction effects, which give a hint for multicollinearity. Hence, a variety of versions to the model are presented to split the correlated variables.
22
Table 5
Cross-sectional regression.
The table shows the results to determine the influence of politically engaged firms on the level of TARP support by using cross-sectional data of all 294 firms that received TARP funds. To test this influence, a OLS regression according to Blau et al. (2013) has been estimated: TARPDUMi = β0 + β1Pricei + β2ln(Sizei) + β3ln(TotAssetsi) + β4D/Ei + β5Turni + β6Volti + β7DUMMYi + εi.The
dependent variable TARPmoney indicates the amount of dollars (in millions) received by firm i as part of TARP. The independent variables include Price, ln(Size), ln(TotAssets), D/E, Turn, and Volt, which we have been defined in section 3. The dummy variable
PACDUM indicates firm i with one if positive lobbying contributions are identified during 5 years prior to their receipt of bailout
funds, and zero otherwise. The variable CONNECT denotes firm i with one if a political connection according to CRP is reported and
PACDUM x CONNECT determines the interaction of PACDUM and CONNECT. Due to the fact that multicollinearity might be an
issue, a variety of versions to the model are presented to split the correlated variables. The standard errors (in brackets) are heteroskedasticity consistent and at a scale of hundreds. Significance levels are indicated: * = 10%, ** = 5%, *** = 1%.
23
Panel A includes the dummy variable PACDUM. In all columns from [1] to [4] a positive relation is detected between firms that lobbied and the amount of TARP money received at a significance level of 0.05. For example, the full model in column [1] shows an estimate of 1,029.5310 and a p-value of 0.032. In economic terms, the estimate PACDUM in column [1] reveals that lobbying firms received $1.03 billion more in TARP support than firms without spending political expenditures.
On average lobbying firms received $1.54 billion more in TARP support than firms that did not lobby. As already detected in the previous subsections 4.1 and 4.2, financial size (Size and TotAssets) seems to matter. A positive relation between financial size and the amount of TARP can be identified at a 0.01 significance level. Furthermore, in columns [1], [3] and [4] and in the remaining panels B and C statistically significant evidence at a 0.05 level is found that firms with lower share turnover and higher idiosyncratic volatility received a greater amount of TARP money. The dispersion for column [2] while controlling for the variable Price can be explained by a non-normal distribution (skewness = 12.67; kurtosis = 182.05). But if the variable Price in the model in column [2] is omitted the results are identical to those reported. On average, lobbying firms received $1.54 billion more in TARP support than firms that did not lobby.
Panel B includes the indicator variable CONNECT. The variable CONNECT reveals significant results at the 0.01 level and positive estimates. On average, political connected firms received $1.59 billion more in TARP support than non-connected firms.
Next, the two indicator variables PACDUM and CONNECT are combined to identify a possible interaction effect. The results from this combination are reported in Panel C. Firms that both lobby and maintain political connections according to CRP receive on average $1.91 billion more TARP money at a significance level of 0.05. The variable PACDUM shows significant but negative estimates, which can be explained by a negative correlation of ρ~-0.93 (unreported). The variable CONNECT is positive and highly significant at the 0.01 level.
24
To follow this aspect, the focus is set on the relation between the amount spent for lobbying and the level of TARP money received. The following model, in line with Blau et al. (2013), has been estimated:
TARPmoneyi = β0 + β1Pricei + β2ln(Sizei) + β3ln(TotAssetsi) + β4D/Ei + β5Turni + β6Volti + β7PACDoli +
β7Connecti + β7PACDol x Connecti + εi (4)
Again, the dependent variable TARPmoney indicates the amount of dollars (in millions) received by firm i as part of TARP. The difference to equation (2) is that PACDol is included instead of PACDUM. The continuous variable PACDol indicates lobbying expenditures and contributions to PACs of firm i. Like before, a multivariate OLS regression approach is used. The results using cross-sectional data are presented in Table 6. The results show evidence of heteroscedasticity and are therefore corrected with White’s (1980) correction, even though similar results are identified (unreported). Evidence of multicollinearity is detected in unreported tests in which the variables Size and TotAssets reveal a high correlation (ρ~-0.45) and show significant interaction effects. Hence, a variety of versions to the model are presented to split the correlated variables. As stated before, variance inflation factors of below 2 are only obtained while omitting either Size or TotAssets. Since Faccio at al. (2006) argue that total assets show a higher economic contribution the variable Size is excluded from the model.
Columns [1] to [4] in Table 6 include the variable PACDol. The estimate is positive and significant (p-value = 0.000) in all columns. The PACDol estimate averages 396.32. Inferred in economic terms the estimate suggests that for every dollar spent on lobbying, firms received on average $396.32 TARP money. As before, evidence in columns [1] and [3] to [8] is found that firms with high D/E ratios, lower share turnover, and higher idiosyncratic volatility received greater amount of TARP money. Again, the estimates of financial size (Size and TotAssets) are positive and statistically significant (p-value = 0.000).
Columns [5] to [8] reveal very similar results while including the interaction variable PACDol x
Connect. The full model in column [5] shows that the interaction estimate is 332.45 and statistically
significant at a 0.01 level. The same observation holds for the remaining columns [6] to [8] as well. On average, firms received $346.77 for every dollar spent on lobbying activities and being connected according to the Center for Responsive Politics.
25
Taking into account all results reported and described in this subsection the third hypothesis that politically engaged firms have an influence on the level of TARP support is confirmed. These results are in line with the original study of Blau et al. (2013) who find for instance that for every dollar spent on lobbying firms received on average $535.71 in TARP funds. The difference might occur due to different data sets and the fact that campaign contributions (PACs) are in general lower than lobbying donations. Again, financial size has an influence on the political engagement factors. This fact is revealed when looking at the correlation between financial size and the lobbying factor
Table 6
Cross-sectional regression.
The table shows the results to determine the influence of politically engaged firms on the level of TARP support by using cross-sectional data of all 294 firms that received TARP funds. To test this influence, a OLS regression according to Blau et al. (2013) has been estimated: TARPDUMi = β0 + β1Pricei + β2ln(Sizei) +
β3ln(TotAssetsi) + β4D/Ei + β5Turni + β6Volti + β7DUMMYi + εi. The dependent variable TARPmoney indicates the amount of dollars (in millions) received by firm i as
part of TARP. The independent variables include Price, ln(Size), ln(TotAssets), D/E, Turn, and Volt, which we have been defined in section 3. The continuous variable
PACDol indicates lobbying expenditures and contributions to PACs of firm i. The variable CONNECT denotes firm i with one if a political connection according to CRP is
4.4. Robustness
First, the results in Tables 5 and 6 are likely underlying a sample selection bias. The sample selection bias is grounded on the assumption that some firms do not find it desirable to be active in political engagements and therefore bias the random sample selection required for consistent estimation. Second, endogeneity is a problem. Maybe the results are driven by other causalities. In other words, the distribution process of TARP funds by the U.S. Treasury department might be affected by other variables or causalities than political engagement or financial size. To test the severity of both problems question (3) and (4) the Heckman correction is applied. The Heckman correction is constructed by a two-stage approach. The first stage is a limited dependent variable estimation of the conditional probability that firm I lobbied given X. X is a vector of control variables. Following Blau et al. (2013), the first equation is estimated using a probit model:
P(PACDUM = 1|X) = ɸ(Xɳ)
PACDUM has been defined in section 3 and ɸ is the standard normal cumulative distribution function. The vector X indicates Price, Size, TotAssets, D/E, Turn, and Volt.
Within the next stage these predicted probabilities are transformed to estimate a model that is close related to equation (3) and (4). Again, following Blau et al. (2013) the two stage Heckman correction equation is estimated as follows:
E[TARP|X,PACDUM = 1] = Xβ + E[ε|X,PACDUM = 1]
The variable TARP has been defined in section 3. Due to evidence of multicollinearity and correlated control variables the Heckman correction is applied in the same array as in Table 5 and 6. The results of the regression using Heckman correction are unreported. They reveal similar results as those reported in Table 5 and 6. For example, the estimate of the variable PACDol in Table 6 is replicated using Heckman correction averages 323.89 and is statistically significant at the 0.01 level. It is inferred that the results in section 4 are robust to sample selection bias and potential endogeneity. Blau et al. (2013) find an oxymoron in the intention of firms to lobby and the distribution process of the U.S. Treasury department. Prior research provides evidence that the motivation for political engagement of firms is based on receiving insurance during periods of economic disruptions (Faccio et al., 2006; Yu and Yu, 2011). The results reported before show a positive relation between political engagement and the likelihood of receiving TARP funds. Hereby an insurance intention is inferred. However, some banks submitted protest against the mandatory acceptance of TARP money14.
14
28
Due to endogeneity, a clear reason why banks contribute in political engagement cannot be identified. Furthermore, as stated in section 2.2, to pay back TARP loans certain U.S. Treasury requirements have been fulfilled. Hence, banks that were forced to accept TARP funds, political engagement seems not to be needed to make use of the “insurance”. Following Blau et al. (2013), the whole multivariate regression is replicated with adding the independent variable Payback. This variable denotes whether a firm had paid back TARP funds by the end of 2009 (the earliest possible repayment date). The unreported results show that the estimate of the variable Payback is not significant and has no impact on other control variables. The investigated relations before are confirmed.
A further test is conducted to illustrate the impact of financial size on the amount of TARP payouts. Therefore, each firm identified with actions in political engagements is compared with a non-engaged firm and those who have the same financial size (market cap or total assets). The magnitude of this comparison is the fraction of TARP funds received. The results show that within the TARP sample the fraction of firms that lobby is 3 times larger than the fraction of firms that did not lobby. These results support the already identified difference in financial size of political engaged firms and non-engaged firms in Table 1.1. The same fraction is calculated for firms that reveal political connections according to CRP. It is found that TARP money received by political connected firms is four times as large as the amount received by non-connected firms.
An additional robustness test is related to the fact that larger firms in terms of financial size appear to receive more TARP money. To test whether financial size (Size and TotAssets) influences the estimate of any political engagement variable (PACDUM, PACDol and CONNECT), the eight largest banks are excluded from the sample. These largest banks were indicated via the list of banks that are considered “too-big-to-fail” according to the Financial Stability Board. After replicating the entire analysis, the same results are identified as discussed above or corroborate them, as it was already done in Table 9 (appendix).
29
5. Conclusion
This study is related to the work of Blau et al. (2013) and examines the relation between political engagement of firms and a beneficial regulatory treatment. The dataset is based on financial firms that were participants in the Troubled Asset Relief Program (TARP) after the financial crisis in 2008, on financial firms that lobbied 5 years prior to the beginning of TARP and on all other public financial firms located in the U.S. as a control group. To investigate this relation three hypotheses have been defined in the introduction:
(1) Political engagement of financial firms leads to a higher probability of receiving TARP support. The reported results in section 4 detect that lobbying firms have a higher probability of receiving TARP funds than non-lobbying firms. The marginal impact on receiving TARP money averages 15% for lobbying firms.
(2) Political engagement of financial firms leads to a more timely payout of TARP money.
The second hypothesis stated is partly confirmed due to the fact that only firms that concurrently lobby and maintain political connections receive TARP money more timely than other firms.
(3) Political engagement of financial firms has an influence on the level of TARP support.
The third hypothesis is confirmed as well. On average political engaged firms received $1.68 billion more in TARP support than politically non-engaged peers. The same test is conducted to find out if the amount spent for lobbying influences the amount of TARP money received. Firms receive on average $358.51 billion for every dollar spent on lobbying during the 5 years prior to TARP in 2008. In general it is inferred that direct and personal political connections have a higher influence within all test frames conducted upon the three hypotheses than lobbying expenditures.
30
The question if financial size leads to a higher probability of investing money in political connections or vice versa might be an interesting topic to investigate further in future research. In other words: Does financial size lead to political contributions or political contribution to financial size? In addition, the distribution process of the U.S. Treasury Authority of TARP money is unknown. As mentioned in the introduction, the U.S. Treasury Department might have used the CAMELS15 rating. Therefore, future research may address this problem and include variables which replicate the CAMELS rating, according to Curry et al. (1999), to investigate the drivers of the TARP distribution process. However, beside the influence of financial size the findings in this study reveal that political engagement was an important determinant in the allocation of TARP money by the U.S. Treasury Department.
Acknowledgments
Due to my research interest in political economy and institutional economics, I already wrote a bachelor thesis to find out whether a regulatory intervention of the German and European government in the banking sector (separate banking system) is to be regarded as economically viable. Therefore, I would like to thank my supervisor Mario Hernandez Tinoco for giving me the opportunity to write my thesis about this topic. I really appreciated his valuable and helpful support.
15
