Government Economic Policy Uncertainty, Corporate Cash Holdings. and the Value of Cash

Transcription

1 Government Economic Policy Uncertainty, Corporate Cash Holdings and the Value of Cash By Hien T. Nguyen, Nam H. Nguyen, Hieu V. Phan, and Shantaram Hegde Hien T. Nguyen, School of Industrial Management, Ho Chi Minh University of Technology, Vietnam National University Ho Chi Minh City, 268 Ly Thuong Kiet Street, District 10, Ho Chi Minh City, Vietnam, phone: (84) ; Nam H. Nguyen, The Manning School of Business, University of Massachusetts Lowell, 1 University Avenue, MA 01854, phone: (978) ; Hieu V. Phan, hieu_phan@uml.edu, Manning School of Business, University of Massachusetts Lowell, 1 University Avenue, Lowell, MA 01854, phone: (978) ; Shantaram Hegde, shegde@business.uconn.edu, School of Business, University of Connecticut, 2100 Hillside Rd, Storrs, CT 06269, phone: (860)

2 Government Economic Policy Uncertainty, Corporate Cash Holdings and the Value of Cash Abstract This research examines the relationships between government economic policy uncertainty and corporate cash holdings and the value of cash. We find robust evidence that policy uncertainty is positively related to corporate cash holdings but negatively related to the value of cash. Our analysis indicates that firms investment delays and precautions induced by policy uncertainty lead to larger corporate cash holdings. Policy uncertainty adversely affects the value of cash, particularly for firms with high growth opportunities, because it discourages firms to use cash for investment in value-increasing projects. Keywords: Policy uncertainty; Cash holdings; Value of cash 1. Introduction Government economic policy uncertainty can have negative effects on the economy. Previous academic research suggests that uncertainty related to government spending, tax, and regulatory and monetary policies exacerbated the Great Recession and slowed the economic recovery (Stock and Watson, 2012; Baker, Bloom, and Davis, 2016). The International Monetary Fund (IMF) and the Federal Open Market Committee (FOMC) came to the same 2

3 conclusion about the harmful effects of policy uncertainty on the economy. 1 In the U.S., the level of policy uncertainty increased significantly during the period , peaking around the failure in raising federal debt-ceiling in August 2011 and the fiscal-cliff crisis at the end of 2012 whereby several previously enacted laws would come into effect simultaneously, leading to an increase in taxes and a decrease in spending. 2 Economic policy uncertainty is suggested to have caused the loss of over one million jobs in the U.S. and more than one percentage point decrease in its real gross domestic product (GDP) during the period (source: Wall Street Journal, April 28, 2013). 3 Given the profound impact of policy uncertainty on the economy, academic researchers have shown increasing interest in investigating the effects of policy uncertainty on corporate policies and firm value. Some recent studies document that government economic policy uncertainty has negative financial and real effects. Gulen and Ion (2016) and Nguyen and Phan (2016) report that firms are more likely to delay investments, particularly those that are irreversible, amid high economic policy uncertainty. Policy uncertainty can increase the cost of external financing, exacerbating firms financial constraints (Pastor and Veronesi, 2013; Gilchrist, Sim, and Zakrajsek, 2014). Panousi and Papanikolaou (2012) report that policy uncertainty increases managers riskaversion. 1 last accessed on May 16, The fiscal cliff was related to the expiration of the Bush tax cuts of 2001 and the planned spending cuts under the Budget Control Act of Available at last accessed on May 22,

4 Cash is an important, liquid corporate asset and is at the discretion of the managers. The secular increasing trend in cash holdings of U.S. firms has attracted increasing interest from investors and academic researchers. Bates, Kahle, and Stulz (2009) report that the average cashto-asset ratio of U.S. industrial firms more than doubled during the period , increasing from 10.5% to 23%. Previous literature offers a number of possible explanations for corporate cash holdings, including transaction costs (Mulligan (1997)), precautionary motives (Opler, Pinkowitz, Stulz, and Williamson, 1999; Bates et al. 2009), corporate governance (Dittmar and Mahrt-Smith, 2007; Harford, Mansi, and Maxwell, 2008), tax incentives (Foley, Hartzell, Titman, and Twite, 2007; Pinkowitz, Stulz, and Williamson, 2013), product market competition (Fresard, 2010), and idiosyncratic risk (Campbell, Lettau, Malkiel, and Xu, 2001). However, no prior research has examined the link between policy uncertainty and corporate cash holdings. Our research fills this gap in the literature by examining the effect of policy uncertainty on corporate cash holdings and its implication for shareholder value. Policy uncertainty can affect corporate cash holdings in several important ways. From the real option perspective, firms are motivated to delay investment amid high uncertainty (Bernanke, 1983; Dixit and Pindyck, 1994; and Gulen and Ion, 2016), leading to larger cash holdings. Larger cash holdings provide more flexibility that allows firms to exploit profitable investment opportunities when they arise in the future. In addition, because policy uncertainty decreases asset returns and increases the cost of external financing, which exacerbate firms financial constraints (Gilchrist et al., 2014; Pastor and Veronesi, 2013; and Brogaad and Detzel, 2015), firms are motivated to increase their cash reserves to buffer against financial shocks and maintain smooth operation. Policy uncertainty can also increase managerial conservatism (Panousi and Papanikolaou, 2012), inducing firms to hold more cash, which is the most liquid 4

5 assets. Given the above discussions, we expect a positive relationship between policy uncertainty and cash holdings. We begin by examining the effect of government economic policy uncertainty on corporate cash holdings. Similar to some recent research related to policy uncertainty (e.g., Panousi and Papanikolaou, 2012; Pastor and Veronesi, 2013; Gilchrist, Sim, and Zakrajsek, 2014; Gulen and Ion, 2016; Nguyen and Phan, 2016), we use the economic policy uncertainty index (labelled BBD index) constructed by Baker et al. (2016) as the measure for government economic policy uncertainty. The BBD index is constructed based on the monthly valueweighted average of three components that include the frequency of newspaper articles containing key terms related to policy matters and economic uncertainty, uncertainty about future changes in the federal tax code measured by the dollar impact of tax provisions set to expire in the near future, and the dispersion in economic forecasts of government spending and consumer price index (CPI) as a proxy for uncertainty about future fiscal and monetary policy. Using a sample that includes 119,322 firm-year observations of 13,981 unique firms over the period , we find that policy uncertainty is positively related to corporate cash holdings. Because this positive relationship could be a result of investment delays and/or firm precautions, and these possible causes are not necessarily mutually exclusive, we perform additional analysis for subgroups of firms sorted on their investment irreversibility. Intuitively, firms with irreversible investments are more likely to delay investments amid high policy uncertainty (Gulen and Ion, 2016), implying that these firms larger cash holdings arise from investment delays. On the other hand, since firms with low or no irreversible investments are less likely to delay investments during periods of high policy uncertainty, a positive relationship between policy uncertainty and cash holdings implies a precautionary motive for cash reserves. 5

6 Our analysis results indicate that policy uncertainty relates positively to both subgroups of firms with high or low investment irreversibility, which indicates that both investment delays and firm precautions explain the positive relationship between policy uncertainty and the level of cash holdings. Since both the BBD index and corporate cash holdings follow an increasing trend over the sample period, one may concern about a possible spurious relationship between the two. Alternatively, policy uncertainty tends to be counter-cyclical, which raises a possibility that the positive relationship between policy uncertainty and corporate cash reserves could simply be driven by business cyclicality. We perform two analyses to address these concerns. In the first analysis, we sort firms into subgroups based on their dependence on government spending and rerun the cash holdings regressions. We find that the positive relationship between policy uncertainty and cash holdings is more pronounced for firms dependent on government spending, implying that government policy uncertainty affects corporate cash holdings. In the second analysis, we sort firms into subgroups depending on whether they belong to pro-cyclical or countercyclical industries. Our results indicate that the positive relationship between corporate cash holdings and policy uncertainty is significant for both subgroups of firms. Taken together, our evidence suggests that the positive relationship between policy uncertainty and corporate cash holdings is unlikely driven by a spurious relationship between the two or business cyclicality. Next, we investigate the relationship between policy uncertainty and the value of cash. As discussed above, as firms delay investments due to policy uncertainty, the undeployed cash can lead to an increase in cash holdings. In addition, policy uncertainty may induce firms to increase cash reserves for precautionary purpose. There are opportunity costs associated with 6

7 cash holdings in these cases because cash is not used for investment in profitable projects that generates shareholder value. This argument suggests that policy uncertainty reduces the value of cash to shareholders. Moreover, to the extent that policy uncertainty negatively affects the value of cash, its effect is expected to be stronger for firms with high growth opportunities because the opportunity cost, which is equivalent to the marginal value generated by the investment of these firms, is higher for firms with high growth opportunities than for firms with low growth opportunities. Our analysis results show evidence, albeit weak, that policy uncertainty has a negative effect on the value of cash for an average firm. However, consistent with our expectation, we find strong evidence that policy uncertainty is negatively related to the value of cash for the subgroup of firms with high growth opportunities. In contrast, we find that policy uncertainty is positively related to the value of cash for firms with low growth opportunities. Our further analysis indicates that the positive relationship between policy uncertainty and the value of cash is concentrated in poorly governed firms, which implies that policy uncertainty imposes an external constraint that can substitute for corporate governance. The BBD index may pick up other non-policy-related economic uncertainty, such as labor market variations, currency uncertainty, stock price volatility, or oil shocks, which implies a potential error-in-measurement problem that could bias the model estimation. Gulen and Ion (2016) point out that the U.S. and Canadian economies are closely related and a shock that affects the economic uncertainty in the U.S. is likely to affect the economic uncertainty in Canada as well. Thus, to address the possible error-in-measurement problem, we estimate the BBD news-based index for the U.S. as a function of the Canadian BBD news-based index and other macroeconomic variables. We then use the residuals from the regression as a proxy for U.S. policy uncertainty in our cash holdings and value of cash models. By construction, the 7

8 residuals are orthogonal to economic uncertainty common to both the U.S. and Canada and other macroeconomic factors included in the model. We rerun the cash holdings and the value of cash regressions with the residuals; however, our findings are qualitatively unchanged. It is possible that firms anticipate policy uncertainty and make decision on cash holdings accordingly. This observation implies reverse causality, another source of endogeneity. To alleviate this endogeneity concern, we use the instrumental variable (IV) regressions with partisan polarization (McCarty, Poole, and Rosenthal, 1997; Poole and Rosenthal, 2000) as an instrument for economic policy uncertainty. The partisan polarization measure tracks legislators ideological position over time. McCarty (2012) argues that partisan polarization makes it more difficult to build legislation, leading to policy gridlock and greater variation in policy. Political polarization is a valid instrument for our analysis because it is directly related to policy uncertainty but there is no obvious reason to argue that political polarization has a direct impact on corporate cash holdings other than through policy uncertainty. The IV regression results are consistent with the OLS results, confirming the robustness of our findings. Some previous research reports significant relationships between political uncertainty and corporate decisions (e.g., Julio and Yook, 2012; Jens 2016), which raise a concern that policy uncertainty may simply pick up the effects of political uncertainty. It is important to note the key differences between policy and political uncertainty. Political uncertainty is typically associated with specific political events such as presidential elections (Julio and Yook, 2012) or gubernatorial elections (Jens, 2016). Gulen and Ion (2016) argue that although elections could be good indicators for uncertainty, they do not tell how policy uncertainty varies during the elections nor do they capture the variation in policy uncertainty between elections. On the other hand, the scope of policy uncertainty is broader and includes different types of uncertainty which 8

9 are directly related to policies. Baker et al. show that the BBD policy uncertainty index spiked around events that are ex ante expected to increase policy uncertainty, such as tight presidential elections, the Gulf Wars, the 9/11 attack, the 2011 debt-ceiling dispute, or the battles over fiscal policy. We are interested in policy uncertainty because it has profound impact on corporate behavior in a continuous manner. However, to ensure that our findings are not confounded by the effects of political uncertainty, we perform additional analyses that explicitly control for political uncertainty proxied by the presidential cycles but our results are qualitatively unchanged. Our research contributes to a burgeoning stream of research that studies the effect of policy uncertainty on corporate behavior and firm value and to a more established stream of literature about the determinants of corporate liquidity. To the best of our knowledge, this research is the first that examines the effects of policy uncertainty on corporate cash holdings, which is one of the most important corporate financial policies, and the value of cash. Our evidence of a stronger negative relationship between policy uncertainty and the value of cash of firms with high growth opportunities highlights a particularly harmful effect of policy uncertainty for the type of firms that tend to create more jobs, which is consistent with earlier evidence that policy uncertainty leads to job loss. The findings of our research are expected to have timely implications for corporate managers, investors, and policy makers given the tremendous policy uncertainty they are facing. The remainder of the paper is organized as follows. We present a description of the data and variables construction in Section 2. Section 3 develops empirical predictions and discusses the research methods and results. Section 4 presents robustness checks and Section 5 concludes the paper. 9

10 2. Samples, Variables Construction, and Descriptive Statistics We obtain U.S. firms accounting data from Compustat, and stock price and returns data from the Center for Research on Security Prices (CRSP) database. The sample period begins in 1985, the first year that the BBD index is available, and ends in Following the literature, we exclude firms from the utility and financial industries from our sample because these firms are highly regulated and their cash holdings may have a different meaning. We use the BBD policy uncertainty index constructed by Baker et al. (2016) as the measure for government economic policy uncertainty. The BBD index is constructed based on the monthly value-weighted average of three components that include the frequency of newspaper articles containing key terms related to policy matters and economic uncertainty, uncertainty about future changes in the federal tax codes measured by the dollar impact of tax provisions set to expire in the near future, and the dispersion in economic forecasts of government spending and consumer price index (CPI) as a proxy for uncertainty about future fiscal and monetary policy. The weights for the news-based, expiration tax code and forecaster disagreement components are 1/2, 1/6, and 1/3, respectively. In our analysis, we construct the policy uncertainty measure as the natural logarithm of the average of monthly BBD index values in a given year. The first component of the BBD index, the news-based uncertainty, captures the intensity of concerns about policy uncertainty. The news-based uncertainty is obtained from 10 large newspapers including USA Today, the Miami Herald, the Chicago Tribune, the Washington Post, the Los Angeles Times, the Boston Globe, the San Francisco Chronicle, the Dallas Morning News, the New York Times, and the Wall Street Journal. An article will be counted if it contains terms in all three categories related to uncertainty, the economy, and policy including: uncertainty or uncertain, one of the terms economic or economy, and one of the terms 10

11 congress, legislation, white house, regulation, federal reserve, or deficit. Baker et al. (2016) find that the news-based uncertainty increases with intense news coverage of events such as stock market crash (Black Monday), the Gulf Wars, terrorist attack (9/11), the Lehman Brothers bankruptcy, and the 2011 debt-ceiling dispute. Baker et al. (2016) estimate the level of uncertainty related to future changes to the tax code by the discounted value of the revenue effects of all tax provisions set to expire over the next ten years. The federal tax code provision expiration data are acquired from the Congressional Budget Office. Finally, these authors estimate the inflation and government purchase dispersion by computing the average of the interquartile ranges of Consumer Price Index (CPI) and federal, state and local governments spending forecasts. The fiscal and monetary policies data are obtained from the Federal Reserve Bank of Philadelphia's Survey of Professional Forecasters. We report the descriptive statistics of the samples for corporate cash holdings and the value of cash models in Table 1. The cash holdings sample consists of 119,322 firm-year observations of 13,981 unique firms, and the excess return sample has 68,787 firm-year observations of 9,248 unique firms. The difference in sample size is due to missing stock returns in CRSP. Cash/Asset is the ratio of cash to the book value of assets (i.e., Compustat items CH/AT). Cash/Net asset is the ratio of cash to net assets, where net assets are defined as the book value of assets minus cash. Policy uncertainty is the natural logarithm of the average monthly BBD index in a given year. Size is measured as the natural logarithm of the book value of assets. Market-to-book ratio is defined as the market value of assets divided by the book assets. Cash flow is calculated as the ratio of earnings after interest, dividends, and taxes but before depreciation to the book value of assets. Net working capital is the ratio working capital without 11

12 cash to the book value of assets. Rd is measured as the ratio of R&D expenditure to sales. Capex is calculated as corporate capital expenditure divided by the book value of assets. Leverage is the ratio of the book value of debt to the book value of assets. Dividend is an indicator variable that takes a value of 1 if a firm pays a common dividend in a given year, and 0 otherwise. Acquisition is defined as a ratio of corporate acquisition expenditure to the book value of assets. Industry sigma is estimated as the average of the standard deviation of the ratio of cash flow to book assets over 10 years for firms in the same 2-digit SIC code industries. Excess return is the difference between a firm s annual stock return in a given year and its benchmark return in the same year where the benchmark return comes from the Fama-French 25 size and book-to-market matched portfolio. Other variables are defined in Appendix A. The average ratio of cash to assets (net assets) is 14.04% (30.20%). The average (median) value of policy uncertainty index is (104.04), which is similar to those reported by Baker et al. (2016) and Gulen and Ion (2016). On average, the research and development expenses, capital expenditure, and acquisition expenditures are 8.27%, 6.4% and 2.2% of the firms book value of assets. The average (median) of excess return for the sample firms is -2.6% (-8.8%). [Insert Table 1 about here] 3. Empirical Predictions, Research Methods, Results and Discussions 3.1. Empirical Predictions Previous research reports that firms delay investments amid high policy uncertainty (Gulen and Ion, 2016; Nguyen and Phan, 2016), which might lead to larger cash holdings. Obviously, firms can distribute cash that is not used for investment to investors but they would consider the trade-off between reducing the unprofitable cash holdings and costly external 12

13 financing due to market frictions if they have to raise external funds to support investments in the future when policy uncertainty decreases. Because policy uncertainty is unlikely to be permanent, holding cash could be more cost effective than paying out first and raising external funds later. In addition, policy uncertainty reduces asset returns and increases the cost of external financing, which exacerbate firms financial constraints (Pastor and Veronesi, 2013; Gilchrist et al., 2014; Brogaad and Detzel, 2015). Facing possible external financing uncertainty and higher costs of capital amid high policy uncertainty, firms are more likely to increase their cash reserves to buffer against financial shocks and maintain smooth operation. Policy uncertainty can also increase managerial conservatism (Panousi and Papanikolaou, 2012). Among different classes of assets, cash is the most liquid one, which could be the asset of choice for conservative managers during periods of high policy uncertainty. Given the foregoing discussion, we state our first testable hypothesis as follows: H1: Policy uncertainty is positively related to corporate cash holdings. There are opportunity costs associated with the increase in cash holdings due to investment delays or for precautionary purpose because cash is not used for investment that generates value for shareholders. To the extent that policy uncertainty negatively affects the value of cash, its effect is expected to be more pronounced for firms with high growth opportunities because their opportunity costs, which are equivalent to the marginal value generated by the investment of these firms, are higher than those of firms with low growth opportunities. This line of argument suggests our second hypothesis as follows: H2: Policy uncertainty is negatively related to the value of cash to shareholders and this relationship is stronger for firms with higher growth opportunities. 13

14 3.2. Empirical Models, Results, and Discussions Policy Uncertainty and Corporate Cash Holdings Our cash holdings model is motivated by Bates et al. (2009) and has the following form: C i,t NA i,t = β 0 + β 1 Policy uncertainty i,t 1 + β 2 Size i,t + β 3 MB i,t + β 4 CF i,t NA i,t + β 5 NWC i,t NA i,t + β 6 Capex i,t NA i,t + β 7 Leverage i,t + β 8 Industry sigma + β 9 Dividend dummy i,t + β 10 R&D i,t Sales i,t + β 11 M&A i,t NA i,t + e i,t, (1) In equation (1), the dependent variable is cash ratio, measured as the ratio of cash to either book value of assets or net assets, where net assets are the book value of assets minus cash and marketable securities. Our test variable is Policy uncertainty, which is proxied by the BBD index. We use policy uncertainty lagged by one period to alleviate endogeneity concern. Consistent with our hypothesis 1, we expect the coefficient β 1 to be positive and statistically significant. We control for several factors which are documented in the literature as having power to explain corporate cash holdings, including firm size, growth opportunities, cash flows, net working capital, capital expenditures, leverage, industry cash flow volatility, R&D investment, and acquisition activities. The model further includes firm fixed effects to control for unobserved time-invariant firm characteristics or industry fixed effects to control for industrywide common factors. 4 Table 2 reports the results of the corporate cash holdings regressions. Because the results are qualitatively similar when the dependent variable is the ratio of cash-to-assets or cash-to-net assets, we report the cash-to-assets regression results for discussion (but the results for cash-to- 4 Similar to Gulen and Ion (2016), we do not control for year fixed effects because the BBD index is the same for all firms in a given year. 14

15 net assets regressions are available from the authors). Consistent with our expectation, the estimated coefficients of policy uncertainty are positive, ranging from to 0.016, and statistically significant at the 1% level. Using the coefficient estimates to calculate the economic effect of policy uncertainty on corporate cash holdings, we find that, holding other variables fixed at their sample means, a one standard deviation increase in policy uncertainty above its mean is associated with 0.39 to 0.64 percentage point increase in corporate cash holdings, which is economically important. [Insert Table 2 about here] Policy uncertainty is expected to have stronger effects for those firms that are more prone to the components of policy uncertainty, such as government spending. To explore this proposition, we use the data on government contracts, which are available from year 2000 onward, to estimate firm sensitivity to government spending. 5 Due to the large government contract data set and the lack of common identifiers between the government contractors and Compustat firms, we use the fuzzy matching method and company names to match government contractors with Compustat firms. We then sort firms into two subgroups based on whether their sales to the government in a given year exceed $1 million. We rerun the corporate cash holdings model separately for the two subgroups and report the results in Table 3. The results indicate that the magnitude of the coefficient of policy uncertainty for the subgroup of firms with sales to the government exceeding $1 million is larger than those for the subgroup of firms with low or no sales to the government. This evidence suggests that the positive effect of policy uncertainty on corporate cash holdings is more pronounced for firms that are more prone to uncertainty related to government spending. 5 Data on government contracts are publicly available at 15

16 [Insert Table 3 about here] Since policy uncertainty tends to be counter-cyclical, one may argue that the positive relationship between policy uncertainty and cash reserves could be driven by business cyclicality. To explore this possibility, in the next analysis, we identify countercyclical and procyclical industries based on the asset liquidation values proxied by firms sales cyclicality (Shleifer and Vishny, 1992; Sharpe, 1994; Almeida and Campello, 2007). We estimate the coefficients of the correlation between a firm s sales and the annual gross national product (GNP) over our sample period, and calculate industry-level correlation coefficients as the average of the correlation coefficients of the firms in the same 2-digit SIC industry. We sort industries into the pro-cyclical (countercyclical) subgroup if their correlation coefficients are above (below) the sample median. Then we rerun the corporate cash holdings models separately for firms in each subgroup. If policy uncertainty captures other effects not captured by business cyclicality, we expect the positive relationship between policy uncertainty and corporate cash reserves to hold for firms in both pro-cyclical and countercyclical industries. The estimation results reported in Table 4 indicate that policy uncertainty has a positive effect on the level of cash holdings of firms in both industry subgroups. [Insert Table 4 about here] The positive relationship between cash holdings and policy uncertainty could be due to investment delays and/or firm precautions. While these reasons for cash holdings are not necessarily mutually exclusive, it is important to determine whether the relationship is also driven by firm precautions. Therefore, we investigate the relationship between policy uncertainty and corporate cash holdings for firms sorted on their investment irreversibility. The intuition is 16

17 that firms with irreversible investments are more likely to delay investments amid high policy uncertainty (Gulen and Ion, 2016), implying that these firms larger cash holdings are likely explained by investment delays. On the other hand, firms with low or no irreversible investments are less likely to delay investment during periods of high policy uncertainty, hence a positive relationship between policy uncertainty and cash holdings suggests that these firms increase cash reserves for precautionary purpose. We estimate the corporate cash holdings model separately for the subgroups of firms sorted on their measures of investment irreversibility. The first proxy for investment irreversibility is capital intensity, which is measured as the ratio of the net fixed assets to the book value of assets. A higher capital intensity measure indicates a higher level of investment irreversibility. We sort firms into the high investment irreversibility (low investment irreversibility) subgroup if their capital intensity is above (below) the sample median. We employ industry redeployability scores as the second proxy for irreversible investments using information from the 1997 capital flow table from the Bureau of Economic Analysis (BEA; Kim and Kung (2013)). The table provides information about the capital expenditures of 123 industries, sorted into 180 asset categories. The redeployability score for each asset category is the ratio of the number of industries using that asset category. An industry s redeployability score is the value-weighted average of the redeployability scores for each asset category in which an industry invests, whereas the weight for each asset category is its share in the industry s total capital expenditures. 6 By construction, a higher industry redeployability score means a lower level of investment irreversibility. We sort firms into the 6 The Bureau of Economic Analysis uses the North American Industry System (NAICS) industries code so we merge the data with our sample using the two-digit NAICS code. 17

18 high (low) investment irreversibility subgroup if their industry redeployability scores are below (above) the sample median. Our third proxy for investment irreversibility is the industry-level sunk-cost measure (Kessides, 1990; Farinas and Ruano, 2005; Gulen and Ion, 2016). To calculate this measure, we first compute the ratios of firms rent expense, depreciation expense and the past three years sales of fixed assets, all normalized by the fixed assets at the beginning of the year. We then calculate the industry medians of these three ratios. Finally, the industry-sunk-cost measure is an index that takes a value of 0 for industries with all three ratios above their sample medians, 2 for industries with all three proxies below their sample medians, and 1 otherwise. A higher value of the index indicates a higher level of investment irreversibility. We sort firms into the high (low) investment irreversibility subgroup if their industry-sunk-cost index equals to 2 (0). We rerun the corporate cash holdings regressions separately for the high and low investment irreversibility subgroups based on each of its proxies and report the results in Table 5. The results indicate that the coefficients of policy uncertainty are positive and statistically significant for both the high and low investment irreversibility subgroups across all three investment irreversibility measures, indicating that the positive relationship between policy uncertainty and corporate cash holdings is driven not only by investment delays of firms with high irreversible investments but also by precautionary cash savings of firms with low irreversible investments. [Insert Table 5 about here] In an unreported analysis, we examine the persistence of policy uncertainty s effect on corporate cash holdings. We find that a high level of policy uncertainty in a given year has a positive effect on corporate cash holdings in the next three years before it tapers off. The length 18

19 of policy uncertainty effect on the level of cash appears to be consistent with the investment delay of about three years documented by Gulen and Ion (2016) Policy Uncertainty and the Value of Cash Our results thus far indicate a positive relationship between policy uncertainty and the level of cash holdings. In this section, we investigate the effects of policy uncertainty on the value of cash. To ensure the robustness of the results, we use two regression models: the excess stock returns (Faulkender and Wang, 2006) and market-to-book value models (Bates et al., 2009). The test variable is policy uncertainty. We control for several variables that have power to explain excess stock returns and market-to-book ratio. The excess stock return model has the following form: r i,t R B i,t = γ 0 + γ 1 Policy uncertainty i,t 1 + γ 2 Policy uncertainty i,t 1 C i,t M i,t 1 + γ 3 C i,t M i,t 1 + γ 4 E i,t M i,t 1 + γ 5 NA i,t M i,t 1 + γ 6 RD i,t M i,t 1 + γ 7 γ 10 L i,t + γ 11 NF i,t M i,t 1 + γ 12 C i,t M i,t 1 C i,t M i,t 1 + γ 13 Leverage i,t C i,t M i,t 1 + I i,t M i,t 1 + γ 8 D i,t M i,t 1 + γ 9 C i,t 1 M i,t 1 + firm fixed effects + ε i,t. (2) B In the above model, r i,t is the stock i s return in year t and R i,t is the Fama and French (1993) size and book-to-market matched portfolio return in year t. The independent variable of interest is the interaction between Policy uncertainty and the change in cash holdings (i.e., γ 2 ). This term captures the effect of policy uncertainty on the value of a dollar of incremental cash holdings. Consistent with our argument, we expect γ 2 to be negative and statistically significant. Y denotes the change in Y from year t-1 to t. M is the market value of equity, C is cash, E is earnings before extraordinary items, NA is assets minus cash, RD is research and development 19

20 expenses, I is interest expenses, D is common dividends, L is market leverage, and NF is the sum of net new equity and debt issues. We estimate equation (2) using the full sample and subsamples sorted on firm growth opportunities and report the results in Table 6. In column 1, we find the average effect of incremental cash holdings conditional on policy uncertainty to be statistically insignificant. In columns 2 and 3, we sort firms into the high and low growth subgroups based on their market-tobook ratios, which proxy for growth opportunities, relative to the sample median. We find that the coefficient of the interaction between policy uncertainty and the change in cash is negative ( ) and significant at the 1% level for the subsample of firms with high growth opportunities, whereas the coefficient of the interaction term is positive (0.165) for the subgroup of firms with low growth opportunities. This evidence indicates that policy uncertainty is associated with a decrease (increase) in the value of cash for firms with high (low) growth opportunities. The magnitude of the effects is also important. Our calculation indicates that, holding other variables fixed at their sample means, a one-standard deviation increase in policy uncertainty above its sample mean reduces (increases) the value of one incremental dollar of cash by 8.72 cents (4.38 cents) for shareholders of firms with high (low) growth opportunities. [Insert Table 6 about here] The positive effect of policy uncertainty on the value of cash of firms with low growth opportunities seems counterintuitive and is worth further consideration. Because cash is at the discretion of managers and free cash flow agency problem can destroy the value of cash (Jensen, 1986), policy uncertainty can impose an external constraint that potentially substitutes for corporate governance in monitoring managers. To the extent that policy uncertainty is a 20

21 disciplining factor, we expect the positive effect of policy uncertainty on the value of cash to be stronger (weaker) for poorly (well) governed firms. To explore this possibility, we augment the excess return regressions with corporate governance measures and their interactions with policy uncertainty and the change in corporate cash holdings. We use the GIM index (Gompers, Ishii, Metrick, 2002) or BCF index (Bebchuk, Cohen, Ferrell, 2004) as a proxy for external governance, and blockholder ownership as a proxy for internal governance. Firms with good corporate governance are those with GIM index (Blockholder ownership) below (above) the sample median, whereas firms with poor corporate governance are those with GIM index (Blockholder ownership) above (below) the sample median. We use corporate governance dummies in our analysis with GIM dummy (Blockholder dummy) taking a value of 1 for firms with poor corporate governance, and 0 otherwise. We rerun the excess return model with the three-way interactions among corporate governance dummies or their continuous versions, policy uncertainty, and the change in cash and report the results in Table A1 in the Detachable Appendix. The coefficients of the three-way interactions are positive and highly significant, implying that the positive relationship between policy uncertainty and the value of cash is stronger for poorly governed firms (to save space, the estimation results using the BCF index are not reported but they are available from the authors). Our second model to examine the effect of policy uncertainty on the value of cash has the following form: MV i,t NA i,t = γ 0 + γ 1 Policy uncertainty i,t 1 + γ 2 Policy uncertainty i,t 1 Cash holdings i,t NA i,t + 21

22 γ 3 Cash holdings i,t NA i,t + Other control variables + year dummies + firm fixed effects+ ε i,t. (3) In equation (3), the dependent variable is the market-to-book ratio of firm i in year t. The independent variable of interest is the interaction between Policy uncertainty and cash holdings (i.e., γ 2 ). Following our hypothesis H2, we expect the coefficient γ 2 to be negative and statistically significant. The model controls for the one-year and two-year lag of the change in earnings ( E i,t, E i,t 2 ), R&D expenditures ( RD i,t, RD i,t 2 ), dividends ( D i,t, D i,t 2 ), interest expense ( I i,t, I i,t 2 ), two-year lag of net assets ( NA i,t 2 ) as well as the two-year lead of earnings ( E i,t+2 ), R&D expenses ( RD i,t+2 ), dividends ( D i,t+2 ), interest expense ( I i,t+2 ), net assets ( NA i,t+2 ), and market value ( MV i,t+2 ), all scaled by the net asset values of firm i in year t. We estimate equation (3) and report the results in Table 7. The estimation results show that the coefficients of the interaction between policy uncertainty and cash holdings are negative (-0.177, , and in columns 1, 2, and 3, respectively) and statistically significant at the 1% level, indicating that policy uncertainty has negative effect on the value of cash. This finding is consistent with the result of the excess return regression model, confirming the robustness of our findings. [Insert Table 7 about here] 4. Robustness Checks It is possible that firms anticipate policy uncertainty and make decision on cash holdings accordingly. This observation implies reverse causality, a source of endogeneity. We use the IV regression to address concern about the endogeneity of policy uncertainty that arises from reverse causality. In particular, we use the partisan polarization measure suggested by McCarty, 22

23 Poole, and Rosenthal (1997), Poole and Rosenthal (2000), and Gulen and Ion (2016) as an instrument for policy uncertainty. This measure tracks legislators ideological position over time. McCarty (2012) argues that partisan polarization makes it more difficult to build legislation, leading to policy gridlock and greater variation in policy. In our research context, political polarization should be a valid instrument because it is directly related to policy uncertainty but there is no obvious reason that it has a direct impact on corporate cash holdings other than through policy uncertainty. We report the IV regression results for corporate cash holdings in Table 8. The first-stage results of the IV regression model reported in column 1 of Table 8 indicate that the coefficient on the instrument is positive (0.496) and significant at the 1% level, confirming its relevance. The Wu-Hausman endogeneity test statistic validates our endogeneity concern. The Kleibergen-Paap underidentification test statistic and the Cragg-Donald weak identification test statistic indicate that our selected instrument is relevant. The results of the outcome regression reported in column 2 of Table 8 show that the coefficient of instrumented PU is positive (0.458) and highly significant, indicating that our findings are robust to endogeneity correction. [Insert Table 8 about here] Both policy uncertainty and equity excess returns can be jointly correlated with unobservable investment opportunities, which implies that our excess return model could be subject to the omitted variable problem, another source of endogeneity. To address this endogeneity concern, we estimate the IV regression for excess return model and report the results in Table 9. The estimation results indicates that the coefficients of the interaction between instrumented PU and the change in cash is negative (-0.314) and statistically significant for firms 23

24 with high growth opportunities but positive (0.441) and statistically significant for firms with low growth opportunities. [Insert Table 9 about here] As discussed above, the BBD index may capture the effects of general economic uncertainty that potentially confound our finding of a positive relationship between policy uncertainty and cash holdings. We address this concern by explicitly controlling for several proxies for economic uncertainty suggested by Bloom (2009) in our corporate cash holdings model. First, using the GDP forecast data from the Philadelphia Federal Reserve s Livingston survey, we calculate the coefficient of variation of GDP forecast as a proxy for expected economic growth uncertainty. Second, we estimate the annual cross-sectional standard deviation of firm profit growth as a proxy for future profitability variation, where firm profit growth is defined as the ratio of the change in net income to average sales. Third, we control for the uncertainty of the equity markets proxied by the monthly standard deviation of stock returns and the Chicago Board Options Exchange s VXO index of implied volatility. Finally, we control for an alternative measure of aggregate macroeconomic uncertainty suggested by Jurado, Ludvigson and Ng (2015). We augment our baseline corporate cash holdings model with these five general economic uncertainty proxies but our finding is qualitatively unchanged. To alleviate concern about possible collinearity between policy and economic uncertainty when their proxies are included in the same regression, we use a 2-step regression model to isolate the effects of policy uncertainty from those of economic uncertainty. Specifically, in the first step, we regress the BBD index on the five economic uncertainty variables mentioned above and obtain the residuals. We then rerun the corporate cash holdings model using these residuals 24

25 as proxy for policy uncertainty. The results reported in Table 10 indicate that the coefficient estimates of the policy uncertainty variable are all positive (0.007, and in columns 1, 2, and 3, respectively) and statistically significant at the 1% level. This evidence suggests that policy uncertainty captures the effects on cash holdings that are above and beyond those captured by the general economic uncertainty measures. [Insert Table 10 about here] The BBD index may pick up some other non-policy-related economic uncertainty, such as labor market variations, currency uncertainty or oil shocks, which implies a potential error-inmeasurement problem that could bias the model estimation. It is noteworthy that the U.S. and Canadian economies are closely related and a shock that affects the economic uncertainty in the U.S. is likely to affect the economic uncertainty in Canada as well. Thus, to address the possible error-in-measurement problem, we follow Gulen and Ion (2016) in estimating the BBD index for the U.S. as a function of the Canadian BBD index and other macroeconomic variables, then we use the residuals (labeled RPU) from the regression as a proxy for policy uncertainty in our cash holdings and value models. By construction, the residuals are orthogonal to economic uncertainty common to both the U.S. and Canada and other macroeconomic factors included in the model. We re-estimate the corporate cash holdings model using RPU in place of the BBD index and report the results in Table 11. The results show that the coefficients of RPU are positive (0.010, and in columns 1, 2, and 3, respectively) and significant at the 1% level, implying that our findings are robust to the error-in-measurement correction. [Insert Table 11 about here] 25

26 We also rerun the excess returns model with RPU and report the results in Table 12. The estimation results indicate that the coefficient of the interaction between RPU and the change in cash holdings is negative (-0.421) and highly significant for firms with high growth opportunities but positive and statistically insignificant for firms with low growth opportunities. This evidence is consistent with our earlier finding, indicating that our excess return model results are robust to the error-in-measurement correction. [Insert Table 12 about here] Previous research documents that political uncertainty, which is positively related to national elections, affects corporate decisions and asset prices. Bialkowski, Gottschalk, and Wisniewski (2008) and Boutchkova et al. (2011) report that firms operating in politically related industries tend to experience higher stock return volatility during the presidential election periods. Based on a sample that includes 248 national elections in 48 countries over the period , Julio and Yook (2012) document a negative effect of presidential elections on investments. To rule out the possibility that policy uncertainty simply picks up the effects of political uncertainty, we control for political uncertainty by augmenting the model with an election indicator variable that takes a value of 1 for a presidential election year during the sample period, and 0 otherwise. The estimation results reported in Table 13 indicate that the coefficient of PU remains positive in all three columns and significant at the 1% level. The results also show that the coefficients of presidential election indicator are positive and statistically significant. [Insert Table 13 about here] 26

27 We further rerun the excess return regressions that additionally control for political uncertainty proxied by presidential elections and report the results in Table 14. We find that the effect of policy uncertainty on the value of cash is not statistically different from zero for the full sample. However, the subsample analysis results show that the coefficient of the interaction between PU and the change in cash holdings is negative (-0.315) and significant at the 1% level for firms with high growth opportunities but positive (0.166) and statistically significant for firms with low growth opportunities. The results also reveal that political uncertainty also has a negative effect on the value of cash, particularly for firms with high growth opportunities. Taken together, the evidence indicates that our findings are robust to controlling for political uncertainty. [Insert Table 14 about here] In another robustness check, we run cash holdings regressions with the dependent variable being the ratio of cash and short-term investment to the book value of asset (i.e., Compustat items CHE/AT). The results reported in Table A2 in the Detachable Appendix indicate that the coefficient of policy uncertainty remains positive and highly significant, suggesting that our findings are robust to the alternative measure of corporate cash holdings. The BBD index is constructed based on three components: the news-based uncertainty, the uncertainty of future expiration tax-code, and the disagreement in the monetary and fiscal policy. The effects of the three policy uncertainty components on corporate cash holdings and the value of cash may vary. Thus, in an additional analysis, we run regressions that include all three components and report the results in Table A3 in the Detachable Appendix. The results suggest that corporate cash holdings are positively related to the news-based uncertainty and the tax-related uncertainty components, but negatively associated with the disagreement in the 27