Coinsurance Effect across Divisional Investment Opportunities and the Value of Corporate Cash Holdings Zhenxu Tong * University of Exeter Xinlin Zhu ** University of Exeter First Draft: June 2014 This Draft: December 2014 Abstract We examine how the coinsurance effect across divisional investment opportunities affects the value of corporate cash holdings. We develop three hypotheses based on the financial constraints, the efficient internal capital market, and the agency problems. We find that the coinsurance effect across divisional investment opportunities reduces the value of cash due to the agency problems, reduces the value of cash for financially constrained firms, but increases the value of cash through the efficient internal capital market. We conclude that the results support our three hypotheses, and that they disclose a specific channel through which firm diversification has the value consequence. JEL Classification: G32; G34 Keywords: Coinsurance effect; Firm diversification; Corporate cash holdings * Corresponding author. Address: Xfi Centre for Finance & Investment, School of Business, University of Exeter, Rennes Drive, Exeter EX4 4ST, United Kingdom. Telephone: +44 1392 723155. E-mail address: z.tong@exeter.ac.uk. ** Address: Xfi Centre for Finance & Investment, School of Business, University of Exeter, Rennes Drive, Exeter EX4 4ST, United Kingdom. Telephone: +44 7450207671. E-mail address: xz288@exeter.ac.uk.
1. Introduction The coinsurance effect across divisional cash flows has been well examined in the literature of firm diversification. For example, Lewellen (1971) argues that the coinsurance effect stemmed from the imperfectly correlated cash flows among divisions reduces the bankruptcy risk of a diversified firm and alleviates a firm s financial constraints. Dimitrov and Tice (2006) find that sales growth rates and inventory growth rates drop more for bank-dependent focused firms than for rival segments of bank-dependent diversified firms during recessions, and argue that diversification alleviates credit constraints through imperfectly correlated cash flows. Tong (2012) finds that the coinsurance effect across divisional cash flows is associated with a higher availability of bank lines of credit, and that diversified firms hold a higher fraction of corporate liquidity in the form of bank lines of credit. However, the coinsurance effect associated with firm diversification is not confined to the imperfect correlation across divisional cash flows. Since normally the investment opportunities in the different divisions of a diversified firm do not arrive simultaneously, this creates another type of coinsurance effect stemmed from the imperfectly correlated investment opportunities among divisions. To our knowledge, this is an under-research area, and there is little evidence in the literature on how firm diversification can have a value impact through the coinsurance effect across divisional investment opportunities. In this paper, we examine how the coinsurance effect across divisional investment opportunities affects the value of corporate cash holdings. We select corporate cash holdings as our research setting due to the following reasons. First, since corporate cash holdings are the most liquid asset in a firm, it provides a promising area to examine the hypotheses related to the efficiency of the internal capital allocation. For example, Myers and Raan (1998) argue that more liquid assets can be turned into private benefits at lower costs. Tong (2011) finds that firm diversification reduces the value of corporate cash holdings through agency problems. Second, corporate cash holdings occupy a significant role in the balance sheet. Bates, Kahle and Stulz (2009) find that the average cash to asset ratio for US firms is 23.2% in 2006. The magnitude of 1
cash holdings implies a potentially important channel through which firm diversification can affect firm value. Third, cash holdings are more comparable across firms. One dollar in cash in a diversified firm is physically the same as one dollar in cash in a single-segment firm. This can mitigate the concern related to the comparability as documented in the literature of firm diversification (e.g., Graham, Lemmon and Wolf, 2002), and facilitates the comparison of the value of cash between diversified firms and single-segment firms. We develop three hypotheses based on the financial constraints, the efficient internal capital market, and the agency problems. First, the financial constraints hypothesis predicts that the coinsurance effect across divisional investment opportunities reduces the value of cash for financially constrained firms, because the imperfect correlation among the divisional investment opportunities reduces the amount of the resources needed at a certain point of time to finance all the investment opportunities. Second, the efficient internal capital market hypothesis predicts that for financially constrained firms, the coinsurance effect across divisional investment opportunities increases the value of cash, because the imperfect correlation among the divisional investment opportunities is associated with lower opportunity costs for the efficient transfers of the resources in the internal capital market. Third, the agency hypothesis predicts that the coinsurance effect across divisional investment opportunities reduces the value of cash, because the imperfect correlation among the divisional investment opportunities magnifies the losses brought by the inefficient cross-subsidization in a diversified firm. We use a sample of 10,510 U.S. firms with 77,090 firm-year observations from 1986 to 2010. We use the methodology in Duchin (2010) and construct a measure of the coinsurance effect across divisional investment opportunities. We examine the impact of the coinsurance effect across divisional investment opportunities on the value of cash based on the methodology in Faulkender and Wang (2006). We find that overall the coinsurance effect across divisional investment opportunities reduces the value of cash. We separate the sample into financial constrained and unconstrained firms and find that the coinsurance effect across divisional investment opportunities reduces the value of 2
cash in both groups. We examine the impact of corporate governance, and find that the coinsurance effect across divisional investment opportunities is associated with lower value of cash in the poorly governed firms. We use the methodology in Raan, Servaes and Zingales (2000) and construct a measure of the efficiency of internal transfers. We find that the coinsurance effect across divisional investment opportunities reduces the value of cash through the inefficient crosssubsidization in a diversified firm, but increases the value of cash through the efficient internal capital market. We also examine the impact of the economic recessions, and find a positive impact of the coinsurance effect across divisional investment opportunities on the value of cash during the recessions. We obtain similar results when we use an alternative measure of the unexpected change in cash holdings. We use three econometric methods and find similar results after controlling for the potential endogeneity problem. Our results support the interpretation that the coinsurance effect across divisional investment opportunities reduces the value of cash due to the agency problems, reduces the value of cash for financially constrained firms, but increases the value of cash through the efficient internal capital market. We conclude that the results are consistent with all of our three hypotheses, and that they disclose a specific channel through which firm diversification has the value consequence. Our paper contributes to the literature by identifying an inefficient link between the coinsurance effect across divisional investment opportunities and corporate cash holdings due to the agency problems. The coinsurance effect across divisional investment opportunities is an under-research area in the literature. To our knowledge, only Duchin (2010) examines how this coinsurance effect affects the level of corporate cash holdings, and he concludes that the coinsurance effect across divisional investment opportunities brings forth an efficient link between diversification and corporate liquidity. Our paper differs from Duchin s study in that we find an inefficient link between diversification and corporate liquidity through the coinsurance effect across divisional investment opportunities due to the agency problems. Moreover, we add to the literature by extending the research on the coinsurance effect associated with firm diversification. While the previous papers have focused on the coinsurance 3
effect across divisional cash flows, we focus on the coinsurance effect across divisional investment opportunities. The mechanism is different between these two kinds of coinsurance effect regarding how they can have the value consequence. Previous papers in the literature argue that the coinsurance effect across divisional cash flows can have the value consequence through the financial constraints (e.g., Lewellen, 1971, Dimitrov and Tice, 2006) and through the limited liability (e.g., Leland, 2007). However, the coinsurance effect across divisional investment opportunities can have the value consequence through the financial constraints, the efficient internal capital market, and the agency problems. The difference in the mechanism of the value consequence between the two coinsurance effects implies that we disclose a specific channel through which firm diversification has the value consequence, and this has not been documented in the literature before. More broadly speaking, our paper extends the literature on the debate about the value consequence of firm diversification. The debate has focused on whether firm diversification results in a diversification discount (e.g., Berger and Ofek, 1995; Lang and Stulz, 1994; Campa and Kedia, 2002; Graham, Lemmon and Wolf, 2002) and through which channels the internal capital market can have value consequence (e.g., Shin and Stulz, 1998; Raan, Servaes and Zingales, 2000; Stein, 1997; Khanna and Tice, 2001; Hoechle, Schmid, Walter and Yermack, 2012). Our paper contributes to this literature by adding a new piece of evidence on how firm diversification has the value consequence through the coinsurance effect across divisional investment opportunities. The paper is organized as follows. Section 2 develops the hypotheses. Section 3 describes the data and the methodology. Section 4 presents the results on the relation between the coinsurance effect across divisional investment opportunities and the value of corporate cash holdings. Section 5 concludes the paper. 4
2. Hypotheses We develop three hypotheses about the relation between the coinsurance effect across divisional investment opportunities and the value of cash. 2.1. Financial constraints A financially constrained firm faces a higher cost when raising funds from external capital markets. If investment opportunities in different divisions are highly correlated, a diversified firm will be more likely to encounter multiple investment opportunities in different divisions at the same time. A financially constrained firm will find it difficult to fund all investment opportunities and have to bypass some positive NPV proects. Duchin (2010) argues that the imperfect correlation among the arrival of the investment opportunities in different divisions reduces the precautionary demand for cash holdings and relaxes a firm s financial constraints. Since the value of cash increases with financial constraints (e.g., Myers and Maluf, 1984; Faulkender and Wang, 2006), we expect that the coinsurance effect across divisional investment opportunities reduces the value of cash. Therefore, we have the following hypothesis. Hypothesis 1: The coinsurance effect across divisional investment opportunities reduces the value of cash for financially constrained firms. 2.2. Efficient internal capital market Stein (1997) argues that with the presence of financial constraints, the headquarter of a diversified firm can allocate the resources more efficiently in the internal capital market. We argue that the coinsurance effect across divisional investment opportunities affects the opportunity costs of the internal capital allocation. Lower (higher) coinsurance effect across divisional investment opportunities are associated with higher (lower) opportunity costs. This can affect the value of cash. For example, consider a financially constrained diversified firm with two segments A and B. On one hand, suppose there is perfect correlation among the investment opportunities in the two segments. When good investment opportunities arise in Segment A, good investment 5
opportunities also arise in Segment B at the same time. Let us assume that the investment opportunities in Segment A are better than the ones in Segment B. According to Stein (1997), the headquarter of the diversified firm should move the resources from Segment B to Segment A. However, it implies that some of the good investment proects in Segment B will be forgone. This corresponds to the situation of higher opportunity costs. On the other hand, suppose there is imperfect correlation among the investment opportunities in the two segments A and B. It implies that the investment opportunities in the two segments will not always occur simultaneously. Then it is possible when good investment opportunities arise in Segment A, Segment B has lower investment opportunities at the same time. Similarly, the headquarter of the diversified firm should move the resources from Segment B to Segment A. However, this corresponds to the situation of lower opportunity costs, because now Segment B has lower investment opportunities and it is less costly to forgo the investment proects in Segment B. Therefore, we have the following hypothesis. Hypothesis 2: The coinsurance effect across divisional investment opportunities increases the value of cash through an efficient internal capital market for financially constrained firms. 2.3. Agency problem Raan, Servaes and Zingales (2000) find that firm diversification is associated with agency problems in the form of inefficient cross-subsidization, which implies that weaker segments are funded by the resources transferred from stronger segments. We argue that the coinsurance effect across divisional investment opportunities magnifies the losses brought by the inefficient cross-subsidization in a diversified firm. Lower (higher) coinsurance effect across divisional investment opportunities are associated with fewer (more) losses. This can affect the value of cash. For example, consider a diversified firm with two segments A and B. On one hand, suppose there is perfect correlation among the investment opportunities in the two segments. When Segment A has weak investment opportunities, Segment B also has weak investment 6
opportunities at the same time. Let us assume that the investment opportunities in Segment A are even weaker than the ones in Segment B. According to Raan et al. (2000), resources will be moved from Segment B to Segment A with the presence of the inefficient cross-subsidization. This corresponds to the situation of fewer losses, because now Segment B is also in a poor status. On the other hand, suppose there is imperfect correlation among the investment opportunities in the two segments A and B. It implies that the investment opportunities in the two segments will not always occur simultaneously. Then it is possible that when Segment A has weak investment opportunities, Segment B has good investment opportunities at the same time. Similarly, the resources will be moved from Segment B to Segment A with the presence of the inefficient crosssubsidization. However, this corresponds to the situation of more losses, because now Segment B has good investment opportunities and some of the good investment proects in Segment B will be forgone due to the inefficient cross-subsidization. Therefore, we have the following hypothesis. Hypothesis 3: The coinsurance effect across divisional investment opportunities reduces the value of cash due to the agency problems. 2.4. Summary of the hypotheses We summarize the above three hypotheses in the following table. The impact of the coinsurance effect across divisional investment opportunities on the value of cash Signs of the predictions: Unconstrained Constrained Financial constraints 0 Efficient internal capital market 0 Agency problems Predictions on the marginal value of cash: Unconstrained Constrained Financial constraints =1 >1 Efficient internal capital market =1 >1 Agency problems <1 <1 This table summarizes the predictions of our hypotheses. A plus (minus) sign indicates a positive (negative) impact of coinsurance effect across divisional investment opportunities on the value of cash holdings. A zero sign indicates no relation between the coinsurance effect and the 7
value of cash. We also compare the marginal value of $1 with one, because these three theories make their own predictions about the value of cash. 3. Data and methodology We describe the data and the methodology in this section. 3.1. Data We obtain the segment-level data from the Compustat/Segment database. We use the Compustat/Industrial Annual database as the source for the firm-level data. We obtain stock return data from CRSP. We get business cycle data from the website of the National Bureau of Economic Research (NBER). We obtain the real gross domestic product data from the website of the Federal Reserve Bank. The sample period is from 1986 to 2010. We match the data from these sources and exclude firms with incomplete data. Following the literature in firm diversification (e.g., Berger and Ofek, 1995), we exclude financial service firms and firms with financial service segments (SIC codes between 6000 and 6999). We require that the difference between the sum of segment sales from Compustat/Segment database and the firm s total sales from Compustat/Industrial Annual database is within 1%. We define diversified firms as the firms that have at least two segments with different 4-digit SIC codes. The observations of multi-segment firms with segments in the same 4-digit SIC codes are regarded as single-segment firms. We winsorize the data to reduce the impact of outliers. Our final sample consists of 10,510 firms with 77,090 firm-year observations. The sample contains 17,367 (59,723) firm-year observations for diversified (single-segment) firms. 3.2. Methodology We describe the methodology in this section. 8
3.2.1 Coinsurance effect across divisional investment opportunities We follow Duchin (2010) and construct a measure of the coinsurance effect across divisional investment opportunities. Investment opportunities are measured by Tobin s Q, which is calculated as book assets plus market value of common equity minus common equity and deferred taxes, divided by the replacement value (0.9 book value of assets + 0.1 market value of assets). We use the average Tobin s Q across all stand-alone firms in an industry as a measure of investment opportunities of the segment in a diversified firm. A diversified firm can be viewed as a portfolio of assets, where each segment represents one of the assets. Thus, the Industry Q Volatility in a diversified firm σ(q) in year t is defined as N N ( q) w i w (1) i1 1 where N is the number of segments in a diversified firm. i ( i, i ) is the volatility of investment opportunities of industry i (industry ). We use the average series over prior ten years to calculate the rolling volatilities of the industries. is the correlation between the investment opportunities in industry i and industry. i, w ) i ( w is the weighting of segment i(segment ) in a diversified firm, which is the ratio of the segment s assets to the total asset of the firm. We then construct a measure of the coinsurance effects across divisional investment opportunities due to the imperfect correlation among investment opportunities. We quantify the coinsurance effect by the difference between σ(q) in the equation (1) and a hypothetical measure of volatility that assumes a correlation of 1 (perfect correlation) among the investment opportunities in all the segments. Coinsurance _ Q N N i1 1 N N w w 1 w w (2) i i i1 1 i i, i The variable captures the reduction in the volatility of investment opportunities due to the coinsurance effect. The variable is positive for diversified firms and is zero for single-segment firms. A higher level of Coinsurance_Q indicates a higher coinsurance effect. 9
3.2.2. Coinsurance effect across divisional cash flows We construct a measure the coinsurance effect across divisional cash flows in the similar way. This is to control for the coinsurance effect stemmed from the imperfect correlation among divisional cash flows. Cash flow is measured by earnings less interest and taxes. The Industry CF Volatility in a diversified firm σ(cf) is defined as follows. N N ( CF) w i w (3) i1 1 where N is the number of segments in a diversified firm. ( ) is the volatility of cash flows of industry i (industry ). We use the average series i i, i w ) over prior ten years to calculate the rolling volatilities of the industries. is the correlation between the cash flows in industry i and industry. ( w i, i is the weighting of segment i (segment ) in a diversified firm, which is the ratio of the segment s assets to the total asset of the firm. And we quantify the coinsurance effect by the difference between σ(cf) in the equation (3) and a hypothetical measure of volatility that assumes a correlation of 1 (perfect correlation) among the cash flows in all the segments. Coinsurance _ CF N N i1 1 N N w w 1 w w (4) i i i1 1 i i, i The variable captures the reduction in the cash flow volatility due to the imperfectly correlated divisional cash flows. The variable is positive for diversified firms and is zero for single-segment firms. A higher level of Coinsurance_CF indicates a higher coinsurance effect. 3.2.3. The marginal value of corporate cash holdings We use the method based on Faulkender and Wang (2006) to examine the impact of the coinsurance effect across investment opportunities on the marginal value of cash. We use the following equation. 10
R RB a b b b b 3 5 9 13 15 1 ΔCash MV Cash 1 1 MV b Coinsurance_Q MV 1 NetFinancing 1 b Coinsurance_CF ΔInterestExpenses b2 Coinsurance_Q ΔCash MV b 1 ε 6 b ΔEarnings 10 b MV 14 1 b4 Coinsurance_CF MV ΔDividends 1 b Leverage ΔCash MV 7 1 ΔCash MV b MV 11 ΔCash MV 1 1 1 ΔNetAsset Cash 1 b 8 b 12 ΔR & D MV 1 Leverage where ΔX indicates the change in the variable X of firm i from year t 1 to t. R is the stock return over fiscal year t 1 to t. RB is the stock i's benchmark return over fiscal year t 1 to t. The benchmark portfolio is one of the 25 Fama and French (1993) portfolios formed on size and book-to-market ratio. MV 1 is the market value of equity at year t 1 computed as price times shares outstanding. Cash Holdings is cash and marketable securities at year t. Earnings is earnings before extraordinary items over fiscal year t 1 to t. Net Asset is net assets (total assets cash holdings) at year t. R&D is R&D expenses over fiscal year t 1 to t. Interest Expenses is interest expenses over fiscal year t 1 to t. Dividends is common dividends over fiscal year t 1 to t. Leverage is leverage (debt/ total assets) at year t. Net Financing is net new equity issues (equity issued minus repurchases)+net new debt issues (debt issued minus debt redemptions) over fiscal year t 1 to t. This methodology essentially represents a long-run event study, where the event is an unexpected change in cash and the event window is the whole fiscal year. The dependent variable is the excess return. It is calculated as the stock return for firm i during a fiscal year t less the stock i s benchmark return over the same period. We use the 25 Fama and French (1993) portfolios formed on size and book-to-market ratio as the benchmark. The breakpoints for the 25 portfolios formed on size and BE/ME and the portfolio monthly returns are from Professor Kenneth R. French s web page 1. For each firm-year observation, a firm is grouped into one of the 25 portfolios based on the intersection between size and book-to-market ratio. The benchmark return is the return of the 25 Fama and French portfolio to which the firm belongs during the fiscal year. 1 Website: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html 11
The main independent variable is the unexpected change in cash holdings. We first define the unexpected change in cash holdings as the realized change in cash. Later we use the net change in cash as an alternative measure. Since both dependent and independent variables are scaled by the lagged market value of equity, the coefficient on the change in cash holdings can be interpreted as a measure of the value that investors place on an extra dollar of cash. 3.2.4. The interaction terms To determine the impact of coinsurance effect across divisional investment opportunities, we construct an interaction term ΔCash Coinsurance_Q. The coefficient of this interaction term represents the impact of coinsurance effect across divisional investment opportunities on the marginal value of cash. We also construct an interaction term ΔCash Coinsurance_CF to control for the coinsurance effect across divisional cash flows. In the regressions, we also include Coinsurance_Q and Coinsurance_CF to ensure that the impact on the value of cash is due to the interaction terms, instead of due to the coinsurance effects themselves. 3.2.5. Control variables As in Faulkender and Wang (2006), we control firm-specific characteristics (changes in profitability, investment and financing policy) that may be correlated with both corporate cash holdings and returns. The other variables in the regression include the change in net asset, the change in earnings, the change in R&D expenses, the change in interest expenses, the change in dividends, the lagged cash holdings, leverage, and the net financing during the fiscal year. We follow Faulkender and Wang (2006) and include two interaction terms in the regressions: ΔCash Casht 1 and ΔCash Leverage. Table 1 describes the univariate statistics. The mean of the variable Coinsurance_Q in diversified firms is 0.0338. Since the mean of the variable Industry Q Volatility is 0.2497, it means that the coinsurance effect across divisional investment opportunities on average reduces the volatility in investment opportunities by around 13.5%. 12
4. Results We report our results in this section. We begin by reporting the relation between the coinsurance effects across divisional investment opportunities and the value of cash for the entire sample. We then divided the sample into financially unconstrained and constrained firms. We next investigate the value consequence of coinsurance effects across divisional investment opportunities for well-governed and poorly governed firms. We then examine the impact of the efficiency of internal transfers and economic recessions. Later we conduct robustness checks. 4.1. The value of cash holdings We first examine the marginal value of cash holdings in the entire sample. We present our results in Table 2. In Column 1, the coefficient of ΔCash can be interpreted as the dollar change in excess return for an extra dollar change in cash. To calculate the marginal value of cash holdings of a mean firm, we need to include the interaction terms ΔCash Casht 1 and ΔCash Leverage to calculate the marginal value of cash holdings. Table 1 shows that the mean firm has a lagged cash holdings equivalent to 19.24% of the market capitalization of equity, and the mean leverage ratio is 22.99%. Therefore, the marginal value of cash to shareholders in the mean firm is $1.00 ( $1.163 0.308 19.24% 0.439 22.99%. We conduct the F-test on the null hypothesis that the marginal value of $1 is one and report the p-value in the brackets. The result (p-value=0.75) suggests that the value of an extra dollar is not significantly different from one. In Column 2, we examine the impact of the coinsurance effects across divisional investment opportunities on the value of cash. The coefficient of the interaction term ΔCash Coinsurance_Q is 6.892 (p-value=0.01). In this regression, we also include Coinsurance_CF and ΔCash Coinsurance_CF to control for the impact of the coinsurance effect across divisional cash flows. To calculate the marginal value of cash for single-segment firms, we use the coefficients of terms ΔCash, ΔCash Casht 1 and ΔCash Leverage. The marginal value of an extra dollar to shareholders is $1.08 and is significantly different from one (p-value=0.01). To calculate the 13
marginal value of cash for diversified firms, we use the coefficients of the terms ΔCash, ΔCash Casht 1 and ΔCash Leverage, ΔCash Coinsurance_Q and ΔCash Coinsurance_CF. The marginal value of cash in diversified firms is $0.87 ( 1.249 ( 6.892 0.0338) 3.003 0.0080 ( 0.25019.24%) ( 0.526 22.99%)). The F-test shows that it is significantly different from one (p-value=0.01). The results indicate that an extra dollar in diversified firms is valued 21 cents less than an extra dollar in single-segment firms. The negative coefficient of the interaction term ΔCash Coinsurance_Q is consistent with the agency hypothesis and the financial constraints hypothesis that the coinsurance effect across divisional investment opportunities reduces the value of cash holdings. Since the marginal value of an extra dollar is significantly below one in diversified firms, it is consistent with the agency hypothesis. 4.2. Financial constraints We divide our sample into financially constrained and unconstrained firms. The criteria of financial constraints are as follows. 1. Payout ratio: Fazzari et al. (1988) argue that financially constrained firms have lower payout ratios. We use the ratio of the sum of dividends and stock repurchases to total assets as a measure of financial constraints. For each year in the entire sample, we sort firms according to their payout ratios and a firm is identified as a financially constrained (unconstrained) firm if its payout ratio is less (more) than the mean of the annual payout ratio distribution. 2. Credit ratings: Firms with credit ratings have higher ability to obtain external financing in public debt markets than those without a rating. Following Kashyap et al. (1994), we retrieve data on credit ratings assigned by Standard & Poor s and classify a firm as a financially constrained (unconstrained) firm if its credit ratings are unavailable (available). Table 3 reports the regression results separately for financially unconstrained and constrained firms. In Panel A, under both criteria, the coefficients corresponding to the interaction term ΔCash Coinsurance_Q are negative and significant in both financially constrained and 14
unconstrained firms. This finding is consistent with the agency hypothesis in that the coinsurance effect across investment opportunities reduces the value of cash, regardless of whether the firm is financially constrained or unconstrained. We report the marginal value of cash for financially unconstrained and constrained firms in Panel B. Across the subsamples, we find that the marginal value of cash for diversified firms is smaller than one. It implies that the market places a lower value on the cash in diversified firms due to the agency problems. 2 4.3. Corporate Governance We examine the impact of corporate governance. We employ two measures of corporate governance: (i) the Gindex proposed by Gompers et al. (2003). Gompers et al. construct a corporate governance index with the data on corporate charters of takeover defences from the Investor Responsibility Research Centre (IRRC) database. A higher Gindex indicates more restrictions on shareholder rights, thus corresponding to poorer corporate governance. (ii) the presence of block holders that hold 5% or more of the company s shares (e.g., Cremers and Nair, 2005). We define a firm to be well-governed firm if the firm s Gindex is in the bottom quartile of the sample (Gindex <7) 3, or if the firm has at least one blockholder. We construct a dummy variable called Poor Governance. The variable is 1 if the firm is not well governed, and is 0 if the firm is well governed. We construct an interaction term ΔCash Coinsurance_Q PoorGovernance. The coefficient of this interaction represents the difference in the marginal value of cash between poorly-governed and well-governed diversified firms. Table 4 shows the results. As displayed in Panel A, Column 1 shows the results when we use Gindex as the measure of corporate governance. 4 The coefficient of the interaction term 2 The Panel B of Table 3 shows that an extra dollar is valued at $0.91 (p-value=0.29) for the mean diversified firm when it is financially constrained. The results imply that for a diversified firm that is financially constrained, the positive impact of financial constraints on the value of cash and the negative impact of agency problems offset each other; therefore we observe that the marginal value of cash is insignificantly different from one. 3 These are the firms with fewer restrictions on shareholder rights. 4 Since only a sub-sample of firms are available in the IRRC database (mostly large firms), the number of observations in this column is 11,385. 15
ΔCash Coinsurance_Q PoorGovernance is 4.319 (p-value=0.05), indicating that the coinsurance effect across divisional investment opportunities reduces the value of cash due to the agency problems. Column 2 shows the results using the presence of blockholders as the measure of corporate governance. The coefficient of ΔCash Coinsurance_Q PoorGovernance is 2.841 (p-value=0.01). Panel B shows that when a diversified firm is poorly governed, the marginal value of cash can be as low as $0.59 (p-value=0.01), indicating that there is a large discount on the marginal value of cash through the coinsurance effect across divisional investment opportunities due to agency problems. Therefore the results in Table 4 are consistent with the agency hypothesis. 4.4. Financial constraints and corporate governance In this section we examine how corporate governance affects the relation between the coinsurance effect across divisional investment opportunities and the value of corporate cash holdings for the sub-samples of financially unconstrained and constrained firms. We report the results in Table 5. For brevity the table only shows the results using payout ratio as the criteria for financial constraints. As displayed in Panel A1, among financially unconstrained firms that are well governed, the coefficient of ΔCash Coinsurance_Q is 1.768 (p-value=0.62). The marginal value of cash in single-segment firms is $1.10 (p-value=0.36) while the marginal value of diversified firms is $1.06 (p-value=0.58). Among financially unconstrained firms that are poorly governed, the coefficient of ΔCash Coinsurance_Q is 4.115 (p-value=0.01). The marginal value of cash in single-segment firms is $0.76 (p-value=0.01). We also find that the marginal value of cash in diversified firms is $0.69 (p-value=0.01). The results are consistent with the hypothesis that the coinsurance effect across divisional investment opportunities has a negative (zero) impact on the value of cash among the firms with lower (higher) level of corporate governance for financially unconstrained firms. Among financial constrained firms that are well governed, the coefficient of ΔCash Coinsurance_Q is 13.267 (p-value=0.29). The marginal value of cash in single-segment firms is $1.68 (p-value=0.01) while the marginal value of diversified firms is $1.44 (p- 16
value=0.01). Among financially constrained firms that are poorly governed, the coefficient of ΔCash Coinsurance_Q is 6.661 (p-value=0.38). The marginal value of cash in single-segment firms is $0.86 (p-value=0.11). We also find that the marginal value of cash in diversified firms is $0.75 (p-value=0.02). Therefore, when we divide the subsample of financially constrained firms into two groups based on the measure of corporate governance, we find evidence that agency problems prevail in diversified firms that are poorly governed. We report the results using the presence of block holders as the measure for corporate governance in Panel A2 and find similar results. These findings are consistent with agency hypothesis. 4.5. The efficiency of internal resources transfers In this section we show the results related to the efficiency of the internal capital allocation. Following Raan et al. (2000), we use a measure of cross-divisional transfers. The transfers are measured by the difference between the investment a segment makes in a diversified firm and the investment it would have made had it been on its own. The investment a segment would have made on its own is approximated by the weighted average of the ratio of capital expenditures to assets for single-segment firms in the same industry. We further subtract the industry-adusted ratio of the capital expenditures to assets averaged across the segments of the firm. Thus, crossdivisional transfers are measured as: CAPEX Assets CAPEX ss Asset ss N 1 CAPEX W Assets CAPEX ss Asset ss Where = 1 N denotes segment, ss refers to single-segment firms, W is the ratio of a segment s assets to a firm s total assets, CAPEX is capital expenditures, Asset is the assets. Since we need to measure the efficiency of the internal transfer of funds, we distinguish between efficient transfers and inefficient transfers. We follow the method in Duchin (2010). We first calculate an overall measure of investment opportunities for a diversified firm. It is computed as the asset-weighted average of the investment opportunities of all the segments for a diversified 17
firm 5. We assign a segment to high productivity (low productivity) divisions group if the segment s investment opportunity is higher (lower) than the overall measure of investment opportunities for a diversified firm. For each firm, we sum the transfers made to high productivity and low productivity divisions. We define two variables: Inefficient Transfers and Efficient Transfers. Inefficient (Efficient) Transfers is the sum of the transfers made to low (high) productivity divisions 6. The results in Table 6 are consistent with our conecture. We first show the univariate statistics of the two variables Inefficient Transfers and Efficient Transfers in the Panel A. We show the regressions in the Panel B. The coefficient of ΔCash Coinsurance_Q is 2.069 (pvalue=0.01), indicating that the coinsurance effect across divisional investment opportunities reduces the value of cash in a diversified firm due to the financial constraints. The coefficient of ΔCash Coinsurance_Q Inefficient Transfers is 93.003 (p-value=0.01). It shows that inefficient transfers to low-productivity divisions are negatively related to the marginal value of corporate cash holdings. Moreover, the coefficient of ΔCash Coinsurance_Q Efficient Transfers is 569.978 (p-value=0.02). It shows that efficient transfers to high-productivity divisions have positive effects on the value of cash. As displayed in panel C, the marginal value of cash for a diversified firm that engages in the inefficient cross-subsidization only is $0.85 (p-value=0.01), while the marginal value of cash for a diversified firm that makes value-enhancing reallocations only is $1.68 (p-value=0.03). If a diversified firm makes both efficient and inefficient transfers, it has a marginal value of cash at $1.52 (p-value =0.07). The results are consistent with all of our three hypotheses. 4.6. Economic recessions In this section we investigate the value impact of the coinsurance effect across divisional investment opportunities during the economic recessions. Since economic recessions increase a 5 As mentioned before in the section 3.2.1., we use the average Tobin s Q across all stand-alone firms in an industry as a measure of investment opportunities of the segment in a diversified firm. 6 A diversified firm can (i) make efficient transfers only, or (ii) make inefficient transfers only, or (iii) make both efficient and efficient transfers. 18
firm s financial constraints, we expect that the recessions may affect the relation between the coinsurance effect across divisional investment opportunities and the value of cash through both the financial constraint hypothesis and the efficient internal capital market hypothesis. We construct a dummy variable called Recession. It is 1 if more than two months in a firm s fiscal year are classified as recession period and 0 otherwise. We then construct an interaction term ΔCash Coinsurance_Q Recession. The coefficient of this interaction represents the difference in the impact of coinsurance effect on the marginal value of cash between recession periods and non-recession periods. Table 7 shows the results. The coefficient of ΔCash Coinsurance_Q Recession is 4.131 (pvalue=0.01). It implies that, during recessions, diversified firm are more likely to efficiently allocate internal capitals so that we find a positive impact of the coinsurance effect across divisional investment opportunities on the value of cash during recessions. We find consistent results in terms of the marginal value of cash in Panel B. Therefore, the results in Table 7 are consistent with the efficient internal capital market hypothesis. 4.7. Alternative measure of the unexpected change in cash holdings We follow Faulkender and Wang (2006) and use the net change in cash holdings as an alternative measure of the unexpected change in cash. We define the net change in cash holdings as the realized change in cash holdings minus the average change in cash holdings in the corresponding benchmark portfolio over the same period. Table 8A shows that the coinsurance effect across divisional investment opportunities significantly decreases marginal value of cash. We find similar results as in Table 2. An extra dollar is valued at $1.06 ($0.83) for single-segment (diversified) firms. We repeat the regressions in Table 8B using the net change in cash separately for financially unconstrained and constrained firms. We find similar results as in Table 3. In Table 8C we examine the impact of corporate governance using the net change in cash. We find similar results as in Table 4. We repeat the 19
analysis for other previous tables using the net change in cash (not reported) and find similar results. Therefore, we find consistent results when we use an alternative measure of the unexpected change in cash. 4.8. About the endogeneity We follow Campa and Kedia (2002) and use three econometric measures control for the potential endogeneity problems. First, we adopt Heckman (1979) two-stage model to tackle with the self-selection problem. In the first stage, we run a probit regression for the likelihood that a firm choose to diversify. The dependent variable is 1 if a firm is a diversified firm and 0 if a firm is a single-segment firm. We use similar independent variables suggested by Campa and Kedia (2002). They include firmspecific characteristics, industry characteristics, and macroeconomic variables. Table 9A shows the probit regression. We calculate the Lambda based on the estimates in the probit regression. The Appendix shows the details of the calculation. In the second state, we include the Lambda in the regression. Second, we use the instrumental variables approach. We follow Campa and Kedia (2002), and use the probability of diversification estimated based on the probit model as a generated instrument. In the first stage, we use all the exogenous variables along with the probability of diversifying as explanatory variables in the decision to diversify. In the second stage, we use the fitted value from the first stage as an instrument for the coinsurance effect across divisional investment opportunities. Third, we use the two-way fixed effect estimation to mitigate the omitted variables problem. We include firm fixed effects and year fixed effects in the regression. To estimate the fixed effect regression, we exclude the firms with only one observation in our sample. This reduces the sample size to 75,705 firm-year observations. Table 9B shows the results. As displayed in Panel A, the coefficients of the interaction term ΔCash Coinsurance_Q are negative and significant across three methods. The results are 20
consistent with the interpretation that coinsurance effect across divisional investment opportunities reduces value of cash. In Panel B, the marginal values of cash for diversified firms are lower than one across the three methods. The results are consistent with the findings in Table 2. We repeat the analysis for other previous tables using the three econometric methods (not reported) and find similar results. Therefore, we find consistent results after controlling for the potential endogeneity problems. 5. Conclusion We examine the relation between the coinsurance effect across divisional investment opportunities and the value of corporate cash holdings. We develop three hypotheses based on the financial constraints, the efficient internal capital market and the agency problems. We construct a measure of the coinsurance effect based on the imperfect correlation across divisional investment opportunities. We use the methodology in Faulkender and Wang (2006) to estimate the value of cash. We find that overall the coinsurance effect across divisional investment opportunities reduces the value of cash. We find that the coinsurance effect across divisional investment opportunities reduces the value of cash in both financially unconstrained and constrained firms. We examine the impact of corporate governance, and find that the coinsurance effect across divisional investment opportunities is associated with lower value of cash in the poorly governed firms. Moreover, we find that the coinsurance effect across divisional investment opportunities reduces the value of cash with the presence of inefficient cross-subsidization, but increases the value of cash through the efficient internal capital market. We find consistent results when we examine the impact of economic recessions, when we use an alternative measure of the unexpected change in cash holdings, and when we use three econometric methods to control for the potential endogeneity problem. Therefore, our results are consistent with the interpretation that the coinsurance effect across divisional investment opportunities reduces the value of cash due to the agency problems, reduces the value of cash for financially constrained firms, but increases the value of cash through the 21
efficient internal capital market. The results support all of our three hypotheses, and we disclose a specific channel through which firm diversification has the value consequence. 22
Appendix: The Calculation of Lambda We assume that a firm s decision to diversify can be expressed as follows. D * it Z it Where: D it D it * D it 1, 0, D it Z it it if it if * 0 D it * 0 D it is an unobservable latent variable. is a dummy variable indicating the status of diversification. is a set of variables that affect a firm s decision to diversify. is an error term. We estimate the above equation using a probit model to get the estimates of γ denoted by γe. Lambda is calculated as follows. Lambda it ( ez ( Z e it it ) * D ) it ( ezit) *(1 D 1 ( Z ) e it Where: is the density function of the standard normal distribution. is the cumulative distribution function of the standard normal distribution. it ) 23
References Almeida, H., 2004. The cash flow sensitivity of cash. Journal of Finance 59, 1777 1804. Bates, T., Kahle, K., Stulz, R., 2009. Why do US firms hold so much more cash than they used to? Journal of Finance 64, 1985 2021. Berger, P., Ofek, E., 1995. Diversification s effect on firm value. Journal of Financial Economics 37, 39 65. Campa, J., & Kedia, S., 2002. Explaining the diversification discount. Journal of Finance 57, 1731 1762. Cremers, K., Nair, V., 2005. Governance mechanisms and equity prices. Journal of Finance 60, 2859 2894. Dimitrov, V., Tice, S., 2006. Corporate diversification and credit constraints: real effects across the business cycle. Review of Financial Studies 19, 1465 1498. Duchin, R., 2010. Cash holdings and corporate diversification. Journal of Finance, 65, 955 992. Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3 56. Faulkender, M., Wang, R., 2006. Corporate financial policy and the value of cash. Journal of Finance 61, 1957 1990. Fazzari, S., Hubbard, R., Petersen, B., 1988. Financing constraints and corporate investment. Brookings Papers on Economic Activity 19, 141 195. Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices. Quarterly Journal of Economics 118, 107 155. Graham, J., Lemmon, M., Wolf, J., 2002. Does corporate diversification destroy value? Journal of Finance 57, 695 720. Heckman, J., 1979. Sample selection bias as a specification error. Econometrica 47, 153 161. Hoechle, D., Schmid, M., Walter, I., Yermack, D., 2012. How much of the diversification discount can be explained by poor corporate governance? Journal of Financial Economics 102, 41 60 Kashyap, A., 1994. Credit conditions and the cyclical behavior of inventories. Quarterly Journal of Economics, 109, 565 592. Khanna, N., Tice, S., 2001. The bright side of internal capital markets. Journal of Finance 56, 1489 1528. Lang, L., Stulz, R., 1994. Tobin s q, corporate diversification and firm performance. Journal of Political Economy 102, 1248 1280. 24