Corporate cash holdings and political connections

Specifically, Faccio (2006) shows political connections increase value for a cross-section of international firms. Goldman et al. (2008) find that politically connected board members lead to increased value. Evidence presented by Fisman (2001), Faccio and Parsley (2007), Chen et al. (2009), and Hill et al. (2010) also show results consistent with political connectedness increasing firm value.

Also, Johnson and Mitton (2003) and Charumilind et al. (2006) show that political connections lead to increased access to debt for firms in Malaysia and Thailand, respectively..

Lobbying activity is defined as the practice of attempting to persuade legislators to propose, pass, or defeat legislation or to change existing laws, which could provide gains for special interests. See www.​senate.​gov for a full discussion of lobbying activity.

Recent growth in lobbying expenditures indicates the increasing importance of lobbying to US firms and industry groups as a means to become politically connected. Total lobbying expenditures increased from $1.45 billion to $2.6 billion over the 1998–2006 period, a seventy-nine percent increase. Over this same period the number of registered lobbyists increased dramatically from 10,693 in 1998 to 15,247 in 2006. See the following link for more details: http://​www.​opensecrets.​org/​lobby/​index.​php.

Chung (1999) discusses various aspects associated with corporate lobbying strategies.

We note that the Faccio (2006) definition of political connectedness is only able to categorize US firms as politically connected based on whether a large shareholder or top officer is closely related to a top politician or party, as top-level US government officials are not allowed to be large shareholders or top officers.

Yu and Yu (2008) state that over half of former congressmen or senators become corporate lobbyists.

This difference in magnitude is also noted by Milyo et al. (2000), Anolabehere et al. (2002, 2003), and Yu and Yu (2010).

We emphasize that the focus of our paper concerns the relation between cash and political connections while Caprio et al. (2008) examine the association between cash and political corruption for international firms.

Their results also show that net working capital and capital expenditures have decreased and that research and development expenditures have increased over their sample period.

Also, improved financial performance will reduce financing frictions by reducing default risk. In turn, the demand for cash would be further weakened.

Other studies examine the market value of cash with respect to various corporate issues. Pinkowitz et al. (2006) show political corruption reduces the marginal value of cash for international firms, while Dittmar and Mahrt-Smith (2007) show weakened governance reduces the value of cash for US firms. Klasa et al. (2008) show the value of cash decreases with increased unionization rates. Finally, Tong (2009) shows that the marginal value of cash decreases with firm diversification.

Research using OPSW’s basic framework include Faleye (2004), Ozkan and Ozkan (2004), Foley et al. (2007), Klasa et al. (2008), Harford et al. (2008), and Bates et al. (2009).

We use this approach instead of 2SLS as Hill et al. (2010) show that cash and lobbying activity are influenced by similar factors making it difficult to find an exogenous regressor to use in the first-stage regression. Specifically, the authors show that corporate lobbying activity is related to size, growth opportunities, research and development expenditures, and industry affiliation, each of which appear in the OPSW framework (Eq. 1).

We use a trailing cash flow volatility measure calculated as follows. For observations in 1998 (first usable sample year), we require at least five panel appearances over the 1987–1997 sample period. We extend the estimation window for later years and maintain the five appearance minimum. For example, the estimation window for 2005 is 1987–2004. This methodology yields a time-variant yet backwards looking cash flow volatility without sacrificing too many observations..

Several studies use variations of the framework to value corporate liquidity. Pinkowitz et al. (2006) use the model to value cash held by international firms, while Pinkowitz and Williamson (2007) and Drobetz et al. (2010) use the model to value cash holdings for U.S. firms. Dittmar and Mahrt-Smith (2007) and Fresard and Silva (2010) use the framework to value excess cash holdings for US firms and US cross-listings, respectively. The model is a variant of the valuation framework developed by Fama and French (1998) to value the tax implications of dividends and debt. Lee (2010) uses the Fama and French (1998) model to examine the impact of managerial entrenchment on the market value of dividends.

The marginal effect of cash on firm value is calculated by taking the partial derivative of the regression equation with respect to cash. Since market value and cash are scaled by net assets, β ₁ represents the change in market value for an additional $1 held in cash. Accordingly, β ₇ represents the marginal impact of political connections on the market value of cash.

As shown in Table 4, we condition cash on other lobbying measures.

The database has a few of the nuances. Firms are required to submit good faith estimates, rounded to the nearest $20,000, of all lobby expenses for each six month period. Firms spending less than $10,000 in a six month period do not have to state lobby expenses, and the CRP treats their lobby expenses as $0. Also, the CRP attributes subsidiary lobby expenses to the parent firm. The CRP provides annual lobbying expenses by summing the mid-year and year-end reported lobbying expenditures.

To check for reporting errors, the Senate’s Office of Public Records matches lobby expenses reported by firms to revenues reported by lobbyists. The lobbying data reported by the CRP does not include amounts spent on industry trade association lobbying and so is considered an estimate of direct firm-level lobby expense. Thus, lobbying expenses may be underreported as spending on indirect lobbying via trade associations or other types of organizations are not reported at the firm-level. A benefit of the OPSW determinants of cash framework (Equation (1)) is that it specifies controls for firm size and Fama and French (1998) industry dummies. This is important as 1) larger firms likely pay a greater share of their respective industries’ trade association lobbying expenses and 2) industry trade association lobbying expenses vary by industry as the benefits from lobbying exhibit cross-industry variation. In a similar fashion, Equation (2) accounts for unobserved heterogeneity that may control for funds given to the industry trade association.

We should note that observations are also lost because of the method used to estimate an accurate and backward looking cash flow volatility measure.

For the estimation of Eq. (2), we begin with original, raw sample of merged Compustat and lobbying data. We then delete observations for missing financial statement data needed to estimate Eq. (2). Also, we winsorize the sample based on the levels of the financial characteristics specified in Eq. (2).

Likewise, using the natural logarithm of the cash-to-net assets ratio for multivariate models is consistent with OPSW.

The descriptive statistics for the lobbying variables are similar to those reported by Chen et al. (2009) and Hill et al. (2010).

Unreported pairwise correlations among the control variables do not suggest collinearity problems.

In addition to being consistent with Harford et al. (2008), Petersen (2008) discusses the benefits of accounting for cross-sectional clustering in the standard errors.

Results for model estimated with Newey and West (1987) standard errors, which correct for heteroskedasticity and autocorrelation, yield quantitatively and qualitatively similar results.

Results are available upon request.

The residuals are again insignificant if we over-identify the first-stage regression and include the natural logarithm of sales. We thank an anonymous reviewer for suggesting that we examine simultaneity bias.

Although the model does not appear to exhibit signs of simultaneity bias, we note that all cross-sectional valuation models that rely on accounting information may suffer from endogeneity problems (Gil-Alana et al. (2011).

This restriction amounts to dropping inflation-adjusted annual lobby expenditures in excess of $9.61 M. Note that the data was initially winsorized at the one percent level for both tails of each variable included in Eq. (1). Thus, dropping the observations with the twenty largest lobby expenditures is after winsorizing the data.

A longer panel would help strengthen future inferences with respect to the impact of changes in political party power and the benefits from corporate lobbying. We thank an anonymous reviewer for suggesting this comment.

While the cash-lobbying coefficients are economically large, please note that the ratio of lobby expenditures to net assets is quite small, which provides for meaningful economic interpretations as discussed later.

The standard deviation of ΔLobby _i,t is 0.0003879. Note that this value is less than that reported for Lobby_Ratio _t-1 reported in Table 1, which is attributable to differences in sample sizes. Descriptive statistic results for each sample are available upon request.

Hill et al. (2010) show that corporate lobbying activity is related to size, growth opportunities, research and development expenditures, and industry affiliation. We do not consider research and development expenditures as an exclusion restriction since it already appears in the structural equation (Eq. 2). Although Hill et al. (2010) measure growth opportunities using the market-to-book ratio, we instrument for this variable using sales growth since a variant of the market-to-book ratio is our dependent variable in the second stage regression. Overall, our first stage regression consists of the variables in Eq. (2) (sans the prior and future changes in lobbying activity) along with lagged natural logarithm of sales, lagged growth in sales, and industry dummies.

We thank an anonymous reviewer for suggesting we estimate this model specification.

Note that the data was initially winsorized at the one percent level for both tails of each variable included in Eq. (2). Thus, dropping the observations with the twenty largest lobby expenditures is after winsorizing the data.