Corporate Cash Holdings and Monetary Shocks. Yiling Deng 1 and Haibo Yao 2. Abstract

Transcription

1 Corporate Cash Holdings and Monetary Shocks Yiling Deng 1 and Haibo Yao 2 Abstract This paper examines the impact of monetary shocks on corporate cash holdings. We find evidence that small industrial firms hold onto cash when monetary policy is too tight and large industrial firms do the reverse both in the short-run and in the long-run. Further tests examine whether the long lasting loose monetary policy results in the pileup of corporate cash holdings. The evidence supports the assumption that industrial firms take the long lasting lower interest rate environment to hoard cash to buffer the monetary policy effectiveness. Key words: Cash Holdings; Monetary Shocks; Taylor Rule JEL: G30, G32, E30, E43, E52 1 Robinson Collage of Business School, Georgia State University, ydeng5@gsu.edu 2 Walker L. Cisler College of Business, Northern Michigan University, hyao@nmu.edu 1

2 Industrial firms hold cash for many reasons. John Maynard Keynes (1936) posits three motives: a transaction motive, a precautionary motive, and a speculative motive. The transaction motive for money demand results from the need for liquidity for day-today transactions to bridge the gap between payments and receipts. The precautionary demand for money refers to holding cash to minimize the potential loss arising from a contingency when access to capital markets is costly. Speculative demand for cash refers to holding cash to take advantage of investment opportunities that may arise in the future. Bates et al. (2009) find evidence supporting both the transaction motive and the precautionary motive from firm specific explanations. They also report consistent evidence supporting an increase in cash holdings in the 2000s that cannot be explained by changes in firm characteristics. One way of understanding why U.S. firms have amassed so much cash is to recognize that holding cash provides firms with unexercised option value, giving them financial flexibility in times of heightened uncertainty. 3 Hodrick (2013) cites the example of Google s CFO Patrick Pichette s motivating the company s holding $48.1 billion of cash at the end of 2012 as giving it the strategic ability to pounce. 4 Another example is that Warren Buffet is noted to think of cash held in his portfolio as a call option allowing him to obtain cheap assets at fire sale prices (such as his $5 billion investment in Goldman Sachs in the depths of the financial crisis). 5 Industrial firms choose their optimal cash holdings in response to market challenges including uncertain economic, fiscal, and monetary environment such as the sustainability of historically low 3 Are U.S. Firms Really Holding Too Much Cash? by Laurie Simon Hodrick, Stanford Institute for Economic Policy Research (SIEPR) policy brief, July, Morgan Stanley Technology Conference, February 29,

3 interest rates. These uncertainties create corporate cash flow volatility, resulting in the option value of holding cash. As far as we know, however, none of the prior empirical corporate cash holdings studies define and use monetary policy shocks (or tightening). As measures of monetary tightening, nominal interest rates and changes in nominal rates could prove misleading and could induce perverse results 6 (Fisher, 1930; Friedman, 1968; Mishkin, 1996). Romer and Romer (2004) document that researchers need to minimize the federal funds rate endogeneity problem 7 to specify a true causal link between monetary policy and other economic variables. Prior studies that examine the impact of monetary policy on corporate cash holdings use changes of federal funds rate to measure monetary policy shocks or tightening (Choi and Kim, 2001; Zaman, 2011). None of the cash holdings studies examine the possible relation between persistent loose monetary policy and increasing corporate cash holdings in the 2000s, when monetary policy is specified as too low for too long (Kahn, 2010). Kahn (2010) posits that too long, too low interest rates may contribute to a buildup of financial imbalances, resulting in misallocation of resources (which includes cash holdings). Increasing corporate cash holdings (Bates et al., 2009) possibly means essentially taking money out of circulation, tamping down economic activity and slowing recovery from crises (e.g., Sánchez and Yurdagul, 2013). The impact of persistent too low interest 5 For Warren Buffett, the cash option is priceless, The Globe and Mail, September 24, For instance, the monetary authorities might think they were providing for a steady cost of credit by holding interest rates constant, but if the expected rate of inflation rose, they would really be fostering easier money and credit conditions, while changes in nominal interest rates reflect changes in inflationary expectations. 7 the funds rate often moved endogenously with changes in economic conditions [such as inflation and output gap]. Such endogenous movements may lead to biased estimates of the effects of monetary policy. 3

4 rates on corporate cash holdings could help explain the slow recovery of the economy from the current financial crisis. Furthermore, prior studies have found mixed evidence on the impact of monetary policy tightness on corporate cash holding. Using changes of funds rates to proxy for monetary policy tightness, Choi and Kim (2001) document that when monetary policy is tightened industrial firms initially increase their cash holdings. Zaman (2011), however, finds the contrary using the same monetary policy variable. Bates et al. (2009) find a negative relation between 3-month T-bill yield which is closely linked to the federal funds rate and the transaction demand for cash, but the relation is not significant. There is a significant need to conduct empirical monetary policy-related research to examine the existing theories regarding the relation between corporate cash holdings and monetary policy shocks. In the analysis that follows, we augment Bates et al. s (2009) analysis to include monetary shocks to examine how monetary policy influences corporate cash holdings. We first examine how corporate cash holdings are affected by monetary shocks during We document that in the short run, industrial firms increase their cash holdings when facing positive monetary shocks, or when monetary policy is too tight. Using firm size to proxy for financially constraint, we find that large firms behave differently from small firms: large firms reduce their cash holdings while small firms increase their cash holdings when monetary policy is tight in the short run, and the same relation extends to the long run. Our findings are robust when using different proxies for corporate cash holding, when using different monetary shock specifications, and when using yearly or quarterly data samples. 4

5 While individually these monetary shocks may contain limited information, collectively they potentially provide insight into whether monetary policy contributed to a buildup of industrial cash holding. We provide direct evidence on the contributing role of sustained monetary shocks to increasing cash holdings in the 2000s documented in Bates et al. (2009). We find that industrial firms in the U.S. accumulate their cash holdings in response to sustained negative monetary shocks during this period. The result is robust when we use different specifications of sustained monetary shocks, for different measures of cash holdings, for different estimation models and for both yearly and quarterly data samples. This essay proceeds as follows. In Section 2, we briefly review three main monetary channels and their theoretical predictions. We then discuss the main monetary policy variables we use to measure monetary policy tightness and develop our main hypotheses in Section 3. In Section 4 we discuss our data set and descriptive statistics and Section 5 reports our empirical results. Section 6 concludes. Previous Research and Theoretical Predictions There are three main channels in explaining the impact of monetary policy on cash holdings: interest rate channel, Tobin s q theory, and the credit channel. Interest Rate Channel The interest rate channel regarding the relation between monetary policy and cash holdings is based on Keynes (1936) three distinct motives of demand for holding cash as discussed above. The nominal interest rate is the opportunity cost of holding cash. Keynes (1936) documents an inverse relation between interest rate and the transaction 5

6 and precautionary demand for cash. Keynes (1936) also documents an inverse relation between interest rates and speculative demand for cash when interest rates are expected to rise (fall) if their current levels are low (high). Previous research and evidence supporting the inverse relation between interest rates and the transaction and precautionary demand for cash includes Keynes (1936), Baumol (1952), Tobin (1956), and Miller and Orr (1966). In addition, Baumol (1952) predicts an inverse relation between cash holdings and interest rates. Baumol notes there is a similarity between the problem of managing a cash balance and that of managing an inventory of some physical commodity. The Baumol- Tobin model (Baumol, 1952; and Tobin, 1956) predicts that when an individual receives her income periodically but wishes to make purchases continuously, the optimal strategy of holding cash is inversely related to the square root of the nominal interest rate. Unlike those of individuals, industrial firms cash balance fluctuates irregularly and sometimes unpredictably over time for both operating receipts and expenditures. Miller and Orr (1966) extend Baumol (1952) to incorporate this up and down cash balance movement characteristic of business operations and find that a firm s optimal average cash balance is inversely related to the nominal interest rate and the relation is more sensitive than that of individuals. Bates et al. (2009) predict that firms and financial intermediaries have become more efficient in handling transactions, leading to reduced transactions-based requirements for cash holdings. They also state that the growth in derivative markets and improvements in forecasting and control suggest, all else equal, a lower precautionary demand for cash holdings. Therefore the interest rate channel predicts an inverse relation 6

7 between interest rate and corporate cash holdings while there is no relation between monetary shocks and corporate cash holdings. Tobin s q Theory Tobin s (Tobin, 1969) q theory provides a mechanism through which monetary policy affects the economy through its effects on the valuation of equities. Tobin (1969) defines q as the market value of firms divided by the replacement cost of capital. High q implies that the market value is high relative to the replacement cost of capital, and new plant and equipment capital is cheap relative to the market value of industrial firms. Firms can then issue equity and get a high price for it relative to the cost of the plant and equipment they are buying. We note that a fall in interest rates stemming from expansionary monetary policy would tend to increase the present value of future cash flows, leading to higher Tobin s q. Firms then can issue equity to purchase new investment goods. Bates et al. (2009) note that firms should have more cash immediately after raising capital. Furthermore, previous research also shows that industrial firms in the US increase their aggregate net debt issues (Gertler and Gilchrist, 1993) and reduce their equity issues (Choe et al., 1993) when monetary policy is tight, leading to increasing leverage and therefore reduced cash holdings consistent with the Tobin s Q prediction. The market timing property of leverage discussed in Baker and Wurgler (2002) emphasizes the cumulative outcome of past attempts to time the equity market, which is more of a long term phenomenon. We predict impact of monetary policy on corporate cash holdings as follows: the fall in interest rates stemming from continuous expansionary monetary policy increases Tobin s q, and encourages firms to issue more equity. Industrial firms should have more 7

8 cash immediately after equity issuance. Therefore Tobin s q theory may exhibit a negative relationship between monetary shocks and corporate cash holdings in the long run. Credit Channels Credit channels (Bernanke and Gertler, 1995) emphasize asymmetric information in financial markets associated with costly verification and enforcement of financial contracts. According to the credit channels theory, firms facing more asymmetric information problems could have difficulty raising external capital or face a higher cost of external funds. This would suggest that firms build cash to hedge future funding needs. Two basic channels of monetary transmission arise as a result of asymmetric information problems in credit markets: the narrow credit channel (also known as bank lending channel ) and the broad credit channel (also known as balance-sheet channel ). The broad credit channel stresses the potential impact of changes in monetary policy on borrowers net worth, cash flow and liquid assets, while the narrow credit channel focuses on the effect of monetary policy on the supply of loans by depository institutions. The credit channel literature has examined how monetary policy affects the demand for cash indirectly through the supply of bank loans (e.g. Bernanke and Blinder, 1988), the liability of firms (e.g. Christiano et al., 1996), and the balance sheet of firms (e.g. Gertler and Gilchrist, 1994). Bernanke and Blinder (1988) document that following a monetary shock, corporate income tends to fall more quickly than costs, cash tends to be squeezed during a period of monetary tightening. The effects of the corporate cash squeeze on economic behavior depend largely on firms ability to smooth the drop in cash flows by borrowing. Bernanke and Blinder (1988) emphasize that monetary policy may affect the 8

9 external finance premium by shifting particularly the supply of loans by commercial banks. Christiano et al. (1996) document that net funds raised by the business sector rises for roughly a year after a contractionary monetary policy shock reflecting a deterioration in firms cash flow due to falling sales, initially unchanged level in production etc. The balance sheet theories begin with the idea that capital market imperfections make the spending of certain classes of borrowers depend on their balance sheet positions, owing to the link that arises between collateralizable net worth and the terms of credit. Gertler and Gilchrist (1994) postulate that swings in balance sheets over the cycle amplify swings in spending. Our prediction regarding the impact of monetary policy on corporate cash holdings through the balance-sheet channel works as follows. Tight monetary policy directly weakens borrowers balance sheets either by reducing net cash flows or by declining asset prices, leading to lower net worth. The lower the net worth of industrial firms, the more severe the adverse selection and moral hazard problems are in lending to these firms, the more difficulty industrial firms could have raising external capital. A weaker financial position with smaller net worth increases the conflict of interest with the lender, because the borrower cannot offer enough collateral to guarantee the liabilities she issues, thus resulting in a higher external finance premium. Similarly, a weaker financial position with smaller net worth could exacerbate stockholder-bondholder conflicts. Accordingly, bondholders could choose to protect themselves by requiring covenants that impose minimum liquidity standards or firms could choose to maintain excess liquidity to blunt the effects of tight monetary policy on the cost of debt. This would suggest that industrial firms build cash to hedge future funding needs in response 9

10 to tight monetary policy especially for small (as a proxy for financially constrained) firms. Using a sample of Indian companies, Pandey and Bhat (2007) find that when monetary policy is tight, cost of external funding increases, the information asymmetry between lenders and borrowers increases that forces companies to reduce their dividend payout/or increase retained earnings. One possible implication is the cash holdings increase. The narrow credit or bank lending channel also relies on credit market frictions while banks play a more central role. Because a significant subset of industrial firms relies heavily or exclusively on bank financing, a reduction in loan supply will force those industrial firms to resort to internal financing, like holding more cash for current or future funding use. Expansionary monetary policy, which increases bank reserves and bank deposits, increases the availability of credit, suggesting that industrial firms may reduce their cash holding. As Bernanke and Gertler (1995) point out, the effects of the corporate cash squeeze on economic behavior depend largely on firms ability to smooth the drop in cash flows by borrowing. Firm size could be used to proxy for the access to capital market. The smaller the industrial firms, the more severe the possible adverse selection and moral hazard problems are in lending to these firms. Gertler and Gilchrist (1993, 1994) study the differential impact of a cash squeeze on different types of firms and find striking differences in behavior between large and small firms. Large firms are at least temporarily able to maintain their levels of production and employment in the face of higher interest costs caused by tight monetary policy. Therefore the credit channel predicts that small firms are more sensitive to monetary policy than large firms. 10

11 Overall, credit channel theory predicts that there is a positive relationship between monetary shocks and corporate cash holdings, and large firms are less sensitive to monetary shocks. We summarize the above analysis with different monetary policy transmission channels and the predicted relationship between monetary policy tightness (ease) and corporate cash holdings in Table 1. Previous Findings Former researchers mainly use either interest rates or changes of interest rates to proxy for monetary policy tightness. Choi and Kim (2001) measure the monetary policy by the change in the federal funds rate 8 and find that upon tighter monetary policy, S&P 500 firms initially increase their cash holdings before reducing them, whereas non-s&p firms reduce cash holdings more quickly. Choi and Kim (2001) also include the current value and eight lags of change in federal funds rate to examine the effects of monetary policy over a longer term. Bates et al. (2009) find a negative relation between 3-month T- bill yield which is closely linked to the federal funds rate and the transaction demand for cash, although the relation is not significant. Zaman (2011) uses the change in federal funds rate as a measure of monetary policy change and finds that when monetary policy is tight, industrial firms tend to reduce their cash holdings. Stern and Miller (2004) define policy mistakes as current policy deviations from optimal monetary policy 9 and argue that a material policy mistake would be to allow a significant rate of inflation or deflation, leading to misallocations of resources. Stern and Miller (2004) document that 8 Choi and Kim (2001) also use the negative value of the mix of nonborrowed reserves and one-periodlagged total reserves for their robustness check. 9 Stern and Miller (2004) do not provide a specific formula of optimal monetary policy, instead they discuss the general framework in 11

12 overly tight monetary policy result in a significant rate of deflation. Holding money relative to physical assets becomes increasingly attractive, so corporate cash holdings increase. On the contrary loose monetary policy results in a significant increase in the rate of inflation; holding money relative to physical assets becomes increasingly costly, so that industrial firms reduce their cash holdings. Policy Deviation and Hypotheses Development Policy Deviation Following the academic literature such as Taylor (1998) and Kahn (2010), we use the Taylor rule (Taylor, 1993) to evaluate monetary policy. The general form of the Taylor Rule may be written as: i t = r + π t + α(π t π t ) + β(y t y t ) (1) where i t represents the recommended short-term interest rate, r represents the equilibrium real interest rate, (π t π t ) represents the deviation of the inflation rate (π t ) from its long-run target (π t ), (y t y t ) represents the output gap the level of real GDP (y t ) relative to potential GDP (y t ), and the coefficients α and β represent the policy maker s responsiveness to deviations from the target output and inflation marks. In short, the Taylor Rule prescribes a target Federal Funds rate based on the deviation of inflation and output from long-run means. Taylor (1998) defines policy mistakes as large departures from baseline monetary policy rules. According to his definition, policy mistakes include excessive monetary tightness and excessive monetary ease. Policy mistakes can be measured through use of deviations from the Taylor Rule given in Equation (1). Kahn (2010) uses building such a policy and document three properties of optimal policies. 12

13 the same deviation as an indicator of whether policy is too tight or too easy. Bernanke (2010) also mentions that to address the question whether policy is nevertheless easier than necessary is to compare Federal Reserve policies to the Taylor rule. Stern and Miller (2004) also define material policy mistakes as deviations of current funds rate from a possible optimal monetary policy, which is set to maximize economic efficiency. This deviation also could be used as monetary shocks in the spirit of Romer and Romer (2004). Policy deviations relative to the Taylor Rule s prescribed rate, therefore, can be calculated as: TRDEV t = (i t i t ) (2) where i t is the actual (nominal) target Federal Funds rate at time t and i t is the prescribed Taylor rule rate set according to Equation (1). As Kahn (2010) documents, such [policy] deviations-especially if they are small and temporary-may represent an appropriate and desirable response to unusual economic or financial conditions. Larger and more persistent deviations, however, may contribute to a buildup of financial imbalances. in addition, monetary policy usually takes three to eight quarters to take effect (e.g., Olivei and Tenreyro, 2007; Labonte, 2013). Sustained deviations, therefore, may be a better indicator of policy mistakes. The purpose of this variation is to capture the idea that the cumulative effect of low interest rates over time drives financial imbalances (Kahn, 2010). For this paper we use sustained deviations to help examine whether keeping policy-controlled interest rates too low for too long contributed to the increasing cash holdings for U.S. industrial firms. Kahn (2010) defines the cumulative policy deviation as the sum of Taylor rule deviations from the first period 13

14 up to period t. Different from Kahn (2010), we calculate the cumulative policy deviation on a rolling basis over the most recent (s+1) periods: t TRDEVSUM t = T=t s TRDEV T (3) where TRDEVSUM t is the cumulative policy deviation at time t, and (s+1) < t. We calculate the cumulative policy deviation as the sum of Taylor rule deviations within the recent four periods (s=3), eight periods (s=7), and twelve periods (s=11) for robustness checks. Our definition of the cumulative policy deviation has practical sense especially when using quarterly data for monetary policy takes four, eight or twelve quarters to take effect, as discussed above. Hypotheses Taylor (1999) notes that when monetary policy was too tight, the recovery from the recession was weak and the eventual expansion was slow for several years; when monetary policy was too easy in the late 1960s and 1970s, inflation skyrocketed. Taylor (2007) also points out that large deviations from business-as-usual policy rules are difficult for market participants to deal with and can lead to surprising changes in other responses in the economy. As discussed from Table 1, the interest rate channel and Tobin s q theory predict a negative relationship between policy deviation and corporate cash holdings, while the credit channel predicts a positive relationship. Furthermore, the credit channel also predicts a stronger relationship for large than for small firms. Cumulative policy deviation provides a monetary policy measure to examine the impact of monetary policy on corporate cash holdings in the long run. This test is extremely important for the 2000s which is specified by the long lasting lower interest 14

15 rates (Kahn, 2010). The most commonly cited evidence that monetary policy was too easy during the period from 2002 to 2006, as the actual federal funds rate is below the values implied by the Taylor rule-by about 200 basis points on average over this five-year period (Taylor, 2007; Bernanke, 2010). If the predictions of the interest rate channel and Tobin s q theory hold in the long run, industrial firms will take the low interest rates opportunity to increase their cash holdings gradually, while if the credit channel theory holds, industrial firms will continuously reduce their cash holdings. Bates et al. (2009) report evidence that an increase in cash holdings in the 2000s cannot be explained by changes in firm characteristics 10. Therefore we also make the hypothesis that industrial firms kept increasing their cash holdings when facing persistent loose monetary policy in the 2000s. We summarize different theories and predictions regarding the relation between three monetary policy variables and cash holdings in Table 1. Data and Descriptive Statistics We construct our sample from Compustat and CRSP for the period 1980 to 2007, extending Bates et al. s (2009) sample period for an extra year 11. It is also reasonable for me to use Taylor rule prescriptions to evaluate the appropriateness of monetary policy for this sample period, when the interest rate policy did not experience significant structural changes. The Federal Reserve System has focused on achieving its objectives for growth in the supply of money and credit following the Monetary Control Act 12 of In 10 In Bates et al. (2009) specification, the dummy variable for the 1990s is significantly negative but the dummy variable for the 2000s is significantly positive. The intercept for the 2000s is higher than for the 1980s or the 1990s. 11 My results reported here are consistent with results reported for the sample period 1980 to The Money Control Act of 1980 required the Fed to price its financial services competitively against 15

16 practice researchers and economists use 2008 as a cut-off point to analyze the impacts of monetary policy. For example, Hilsenrath 13 (2013) uses the PFC era and AFC era to refer to the Pre-Financial Crisis and After Financial Crisis. Hilsenrath (2013) also documents that in the PFC era the central bank managed just one short-term interest rate and expected that to be enough to meet its goals for inflation and unemployment. That rate is the federal funds rate Our macro data set includes real GDP data from the Bureau of Economic Analysis, potential real GDP data from Congressional Budget Office, and CPI data from Bureau of Labor Statistics 14. Consistent with the previous literature, we exclude financial firms (SIC codes ) because they may carry cash to meet capital requirements rather than for the economic reasons studied in our analysis. We also exclude utilities (SIC codes ) because their cash holdings can be subject to regulatory supervision. Furthermore, we restrict our sample to firms that are incorporated in the United States to minimize the impact of repatriation tax burdens (Foley et al., 2007). Firm-specific accounting variables are obtained from Compustat, and stock returns are obtained from CRSP. Following Bates et al. (2009), we eliminate firm-years or firmquarters for which book value of total assets is negative or the sales revenue is negative. Our final sample contains 118,897 firm-year observations for 13,743 unique firms and 439,659 firm-quarter observations for 13,210 unique firms. Macro-variables are reported on a calendar year while firm specific variables from Compustat have both fiscal and calendar basis. Most yearly information from private sector providers and to establish reserve requirements for all eligible financial institutions. 13 See Easy-Money Era a Long Game for Fed by Jon Hilsenrath, March 18, 2013, on page A2 in the U.S. edition of The Wall Street Journal. 16

17 Compustat is on a fiscal basis while quarterly information from Compustat has both fiscal and calendar basis 15. For this analysis, we do our research with the same framework as in the finance literature. Specifically, we calculate the average for those quarterly macro variables based on the fiscal year definition for each firm and merge the macro information with firm specific information for the same firm, then combine all information together to get the whole sample. In Compustat, we have the fiscal year end variable (FYR) ranging from 1 to 12. For example, if the fiscal year ending month is January (FYR=01) 1995, then the calendar dates spanned is from Feb. 1 st, 1995 to Jan. 31 st, While for another firm if fiscal ending month is July (FYR=07) 1995, then the calendar dates spanned is from Aug. 1 st, 1994 to Jul. 31 st, Because our macro data are quarterly, we calculate the average of the quarterly values for those four quarters in the calendar year of 1995 for the first case. For the second case we calculate the average of four quarterly values for the third and fourth quarters in the calendar year of 1994 and the first and second quarters in the calendar year of For each of the calculations we ensure that for each fiscal quarter, we have at least 2/3 of the corresponding calendar months. Panel A of Table 3 reports descriptive statistics for variables used in our cash holdings regressions. The variables are defined as follows. Cash: Following Bates et al. (2009), we measure the corporate cash holdings as cash and marketable securities (data item #1) divided by total assets (data item #6). We 14 We follow Kahn (2010). 15 For example, using calendar quarterly Compustat data, Choi and Kim (2005) find that trade credit helps firms absorb the effect of credit contraction i.e. when monetary policy is tight, industrial firms will increase their trade credit. While Haan and Sterken (2006) find evidence that when monetary policy is tight, industrial firms will reduce their trade credit based on fiscal annual accounting data. Although those 17

18 also measure cash holdings as log net cash ratio, defined as log value of cash and marketable securities (data item #1) divided by (total assets (data item #6)-cash and marketable securities (data item #1)), for robustness check. Monetary policy variables: Following Kahn (2010), we take the first two specifications of the Taylor Rule shown in Table 2 to calculate Taylor rule prescriptions in Equation (1) 16. Inflation is measured by the four-quarter rate of change in the CPI and the output gap measured as the log ratio of real GDP to the CBO estimate of potential GDP. Policy deviations are calculated as the difference between the effective funds rates and Taylor rule prescriptions based on Equation (2). Cumulative policy deviation is calculated as the sum of policy deviations from four periods ago based on Equation (3). We include the squared policy deviation, (Policy Dev. t-1 ) 2, to allow for the possibility of a nonlinear relationship between Taylor rule deviations and corporate cash holdings, that is, the possibility that large deviations are much more important than small deviations. As discussed from Section 1, the credit channel theory predicts that industrial firms increase their cash holdings when facing tight monetary shocks because when monetary policy is tight, the assets they use collateral to get external funding depreciate, the problem of asymmetric information becomes more serious. Also there is a big difference between large and small firms when large firms have more and wider access to external funding than small ones, therefore large firms are less sensitive to monetary shocks than small ones with more financing flexibility. Because the prediction is about asymmetric information and market friction, therefore the credit channel prediction researchers use U.S. firms and Euro and UK firms separately. 16 The difference between Type 1 and Type 4 Taylor rule prescriptions is a constant, and the difference between Type 2 and Type 3 Taylor rule prescriptions is also a constant. 18

19 regarding corporate cash holdings and monetary shocks becomes more obvious and significant when monetary policy is tight other than loose. Former literature such as Gertler and Gilchrist (1994) and Choi and Kim (2005) discuss the differences between small and large firms in explaining corporate behavior. Also the piling up of record amounts of cash for large companies attract the attention of monetary policy researchers (e.g., Sánchez and Yurdagul, 2013). We control for unobserved heterogeneity with a dummy variable Large. Specifically for each fiscal year and quarter, we sort all industrial firms in our sample into four quartiles based on firm size. We then define a dummy variable Large equal to one if one firm is in the largest firm size quartile, and zero otherwise. To test the credit channel prediction, we further include a monetary policy interaction variable, Policy deviation t-1 tight large, which is set to check whether large industrial firms are less sensitive to monetary shocks especially when monetary policy is tight than small ones. We also include interaction variables with squared monetary shocks for large firms and when monetary policy is tight to show whether the behavioral difference between large and small firms is nonlinearly different. The two extra interaction variables we include are (Policy Dev. t-1 ) 2 tight and (Policy Dev. t- 1) 2 tight large. To test the interest rate theory, we include the federal funds rate to proxy for the cost of holding cash for industrial firms. The prediction of the Tobin s Q theory is regarding the long-run relation between monetary shocks and corporate cash holdings, for industrial firms take advantage of their stock overpricing as a result of cumulative market timing choice. Therefore, to test the Tobin s Q theory, we include the sustained monetary shocks instead of the temporary 19

20 monetary shocks in the same regression model. As also shown in Figure 2, transient and sustained monetary shocks tell the same story for the pre-2000 period when the Fed overall followed well the Taylor rule prescription, while the 2000s period is specified as the long lasting lower interest rate era which is better specified by the sustained monetary shocks. Furthermore, Bates et al. (2009) report a jump of corporate cash holdings in the U.S. compared with corporate cash holdings in the 1980s and 1990s. To test the Tobin s Q prediction regarding the relation between sustained monetary shocks and corporate cash holdings, especially for the 2000s period when corporate cash holdings experienced a jump, we include the sustained monetary shocks for the last four periods. To show the cumulative effects of the lower interest rates for the 2000s period, we include the interaction variable Cumulative policy deviation 2000s dummy in which 2000s dummy is a dummy variable equal to one if the firm observation is in the fiscal year after 1999, and zero otherwise. We also include two other interaction variables to show whether there is a behavioral difference between small and large firms when monetary policy is continuously loose: Cumulative policy deviation large 2000s dummy and Cumulative policy deviation large. The credit channel predicts that large firms are less sensitive to monetary shocks while the Tobin s Q theory predicts no relation. Macro control variables: Following Bates et al. (2009), we use the credit spread to proxy for the general economic environment such as the default risk and the precautionary demand for cash for industrial firms. Credit spread is the difference between the AAA and BBB yields reported by the Federal Reserve. To control for fiscal policy, we use the fiscal deficit. Although we have the annual federal deficit available for 20

21 each fiscal year from 1930, we cannot find the corresponding quarterly federal deficit data. As a proxy, we use the federal government current receipts and current expenditures data from the U.S. department of Commerce: Bureau of Economic Analysis 17. Specifically, we calculate the quarterly Federal Deficit as the difference between quarterly federal government current receipts and current expenditures divided by nominal quarterly GDP then multiply by 100. To make our analysis consistent, for the annual analysis we calculate the annual Federal Deficit as an average of the quarterly Federal Deficit variable defined above, not the exact annual federal deficit for each fiscal year 18. We do not include inflation and output gap into our analysis because the specification of the theoretical Taylor Rule prescription already includes those two key variables. We include the average effective federal funds rate to proxy for the opportunity cost of holding cash. Firm specific control variables: The control variables in the cash holdings regressions are motivated by the variables used in Bates et al. (2009). Industry sigma is the average across the two-digit SIC code of the firm cash flow standard deviations for the previous 10 years, and we require at least three observations for the calculation. Market-to-book is the ratio of the market value of assets to the book value of assets i.e. book value of assets (#6) minus the book value of equity (#60) plus the market value of equity (#199* #25) as the numerator of the ratio, and the book value of assets (#6) as the denominator. Real size is the logarithm of book assets (#6). Cash flow/assets is calculated as earnings after interest, dividends, and taxes but before depreciation divided by book assets (((#13 #15 #16 #21)/#6). NWC/assets is net working capital (data item #179)

22 minus cash and marketable securities (data item #1) divided by book assets. Capex is the ratio of capital expenditures (data item #128) to the book value of total assets (data item #6). Leverage is the ratio of total debt to the book value of total assets (data item #6), where debt includes long-term debt (data item #9) plus debt in current liabilities (data item #34). R&D/sales is the ratio of research and development expense (data item #46) to sales (data item #12). Dividend dummy is a dummy variable equal to one if the firm paid a common dividend and zero otherwise. Acquisition activity is the ratio of expenditures on acquisitions (data item #129) relative to the book value of total assets (data item #6). Net debt issuance is calculated as annual total debt issuance (data item #111) minus debt retirement (data item #114), divided by the book value of total assets (data item #6). Net equity issuance is calculated as equity sales (data item #108) minus equity purchases (data item #115), divided by the book value of total assets (data item #6). Loss dummy is a dummy variable equal to one if net income (data item #172) is less than zero, and zero otherwise. All variables in dollars are inflation-adjusted to 2007 dollars using the Consumer Price Index. Outliers in a firm-year are winsorized as follows: Leverage is winsorized so that it is between zero and one; R&D/assets, R&D/sales, acquisitions/assets, cash flow volatility, and capital expenditures/assets are winsorized at the 1% level; the bottom tails of NWC/assets and cash flow/assets are winsorized at the 1% level; and the top tail of the market-to-book ratio is winsorized at the 1% level 19. After excluding winsorized and 18 For example, 19 Detailed definitions of those variables were shown in the Appendix. 22

23 missing explanatory values, we am left with 77,738 firm-year observations for 12,430 unique firms, and 218,502 firm-quarter observations for 10,636 unique firms. As reported in Panel A-1 from Table 3, the average annual cash holdings is large at 13.9% of the total assets. The median cash holding, however, is much smaller at 6.7% of the total assets. Taylor rule prescriptions in the sample range from 2.9% to 19.3% for both Taylor rule parameterizations. Taylor rule deviations have means of 0.5% and 0.1%, and medians of 0% and 0.2%, implying that the Fed on average closely follows the Taylor rule prescriptions (e.g. Bernanke, 2010). We report the descriptive statistics for the cumulative policy deviation for the most recent four periods. For the yearly data sample, the cumulative policy deviation is the sum of the current policy deviation and policy deviations within the last three years. Two types of cumulative policy deviations tell different stories about the monetary policy. Type I cumulative policy deviation has an average of -0.2% and median of -0.6%, suggesting that monetary policy is too loose in the long run. On the contrary, Type II cumulative policy deviation has an average of 1.3% and median of 1.3%, suggesting that monetary policy is too tight in the long run. For comparison, we also report the descriptive statistics of our quarterly data sample in Panel A-2 of Table 3. Quarterly medians for both types of policy deviation are zero, consistent with the annual results that the Fed closely follows the Taylor rule prescriptions. Different from the yearly averages, quarterly averages for both types of policy deviation are negative, suggesting possible conflicts between annual and quarterly analysis. Furthermore, both types of quarterly cumulative policy deviation report that monetary policy is relatively loose in the long run, since averages and medians are negative for both. 23

24 Panel B of Table 3 reports Pearson correlation coefficients among those monetary policy variables and macro control variables. As seen in the Panel B-1 of Table 3, cash holdings are negatively related to all four monetary policy variables for both Taylor rule specifications: Taylor rule prescriptions, policy deviations, squared policy deviations and cumulative policy deviations. A number of other noteworthy correlations are evident in the panel. For example, Taylor rule prescriptions are negatively correlated to policy deviations ( for Type I monetary variables and for Type II monetary variables). When the federal funds rate should be set high, the monetary policy is looser than prescribed possibly reflecting the gradualism of the Fed. Cumulative policy deviations are significantly and positively related to policy deviation. The correlation coefficient is for Type I monetary variables and for Type II monetary variables. Quarterly correlation results reported from Panel B-2 in Table 3 support the negative correlation between cash ratio and three monetary variables: Taylor rule prescriptions, policy deviations cumulative policy deviations. But the correlation coefficient between cash ratio and squared policy deviations are significantly positive. Quarterly correlation results also show that cumulative policy deviation is more related to policy deviation for both types of monetary variables on a quarterly basis than on a yearly basis. One must be careful not to draw conclusions from these simple correlations, because Panel B-1 and Panel B-2 in Table 3 also reveal that cash holdings and monetary policy variables are strongly correlated with two control variables: fiscal deficit and credit spread. In Bates et al. s (2009) analysis, the credit spread variable is positive and significant at the 10% level in explaining the formation of cash holdings. Panel B-1 in 24

25 Table 3 reports the correlation coefficient between Type I policy deviation and credit spread is a significant and the correlation coefficient between Type II cumulative policy deviation and fiscal deficit is a significant Quarterly correlation results from Panel B-2 in Table 3 support the above findings. It is possible that when monetary policy is tight, the default risk will increase correspondingly, and also the monetary policy could be set continuously easer to lower the default risk for price stability and economic growth purposes for the Fed. We also find a significant negative relation between credit spread and the fiscal deficit defined above. The correlation is for our quarterly data sample and for our yearly data sample. When default risk is high reflecting a deteriorating economy, federal government current receipts will decrease relative to its current expenditures. This fact is important in understanding the following regression results. Table 4 presents univariate comparisons of key descriptive variables by policy deviation quartiles for both the yearly and quarterly data sample. To show the impacts of tighter or looser monetary policy, we first divide the whole sample into two subsamples: one with negative policy deviations and the other with positive policy deviations. For each subsample we construct four quartiles. We are interested in whether changes of cash or cash ratios will be different for each policy deviation quartile. Panel A-1 and A-2 in Table 4 present sorting results specified by Type 1 Taylor rule deviations. For example, in the yearly sample sorting results from Panel A-1 in Table 4, we find that for the loose monetary policy regimes specified by negative policy deviation quartiles, industrial firms increase their holding of cash and cash equivalents when monetary policy is looser. Industrial firms in the U.S. increase their average cash 25

26 holdings by about million dollars, and median of about million dollars per year in the first quartile, compared with the fourth quartile in which industrial firms reduce their average cash holdings by million dollars. For the other quartiles with positive policy deviations when monetary policy is tighter, industrial firms reduce their cash holdings by an average of million dollars in the fourth quartile, compared with the reduction of cash averaged around million dollars in the first quartile. Panel A- 2 in Table 4 reports quarterly variable changes sorted by quarterly Type I policy deviation. We find that for the negative policy deviation quartiles, industrial firms increase their holding of cash and cash equivalents when monetary policy is looser. Industrial firms in the U.S. increase their average cash holdings by about million dollars, and median of about million dollars per quarter in the first quartile, compared with the fourth quartile in which industrial firms reduce their average cash holdings by million dollars and median of about million dollars. For the quartiles with positive policy deviations, industrial firms reduce their cash holdings by an average of million dollars and median of million dollars in the fourth quartile, compared with the reduction of cash averaged around million dollars in the first quartile. Panel B-1and B-2 present the same results as those from Panel A-1 and A-2, but using Type 2 Taylor rule policy deviations. Panel A-1 to Panel B-2 also report changes of cash ratios sorted by two types of Taylor rule policy deviations. Mean and median changes of cash ratios and total amount of cash holdings at two extreme quartiles seem to support both of our hypotheses, when monetary policy is extremely tight or ease, industrial firms will increase their cash holdings. 26

27 Empirical Results Hu (1999) documents that since monetary policy is more likely to be responsive to macro-level variables than to firm-level variables, the endogeneity problem of monetary policy should not be a cause for concern. Following Hu (1999), we use lagged monetary policy variables in the estimation to minimize the endogeneity problem. We also use both annual and quarterly data samples, different specifications of cumulative policy deviations, and different lags of monetary policy variables for robustness checks. Furthermore, we use different Taylor rule specifications for the monetary policy variables. Table 5 reports regressions of cash holdings on transient monetary shocks and controls for our yearly and quarterly data sample analysis in Panel A and Panel B separately. In both Panel A and Panel B of Table 5, Models (1) to (4) report the firm fixed effects estimation results for cash to total assets ratio for two types of Taylor rule based monetary shocks; Models (5) to (8) report the firm fixed effects estimation results for log value of the cash to net assets ratio for two types of Taylor rule based monetary shocks. Models (1), (2), (5) and (6) report regression results with Type I monetary policy deviation as independent variables, and Models (3), (4), (7) and (8) report regression results with Type II monetary policy deviation as independent variables. Consistent with Bates et al. (2009), we find that as a measure of default risk, credit spread is positively significant supporting the precautionary demand for cash. Consistent with Bates et al. (2009) in predicting that the relation between cash holdings and opportunity cost of holding cash is not significant, for firms and financial intermediaries have become more efficient in handling transactions, thus reducing transactions-based requirements for cash holdings. We find that the federal funds rate, as a measure of opportunity cost of holding 27

28 cash, is not significant and the signs for federal funds rate are mixed for different corporate cash holdings measures. The corresponding implication is that the interest rate channel regarding the relation between monetary policy and corporate cash holdings cannot help explain the increasing cash holdings puzzle. We also find that fiscal deficit is significantly negative for all the firm fixed effects models and for both yearly and quarterly data samples. Our other findings about firm specific explanations for corporate cash holdings are consistent with those from Bates et al. (2009). Models (1), (3), (5) and (7) in Panel A and Panel B of Table 5 report regression results with monetary policy deviation and squared policy deviation as independent variables. Take Model (1) from Panel A for example, the marginal effect of monetary shocks is calculated as ( *Policy Deviation), which means that the change in the corporate cash to total assets ratio is a convex function of the policy deviation. The tighter the monetary policy, the higher corporate cash holdings will be. When monetary policy is tight from neutral i.e. zero policy deviation, a one unit increase of policy deviation holding all other independent variables constant leads to a 14 basis points increase in the corporate cash holdings. And industrial firms increase more when policy deviation becomes wider. Also when monetary policy is too loose (monetary shocks less than -3%), industrial firms will also increase their cash holdings and will increase more when monetary policy is much looser. Therefore, our findings from Models (1), (3), (5) and (7) support the credit channel theory in stating that industrial firms increase their cash holdings when monetary policy is tight and the Tobin s Q theory in stating that industrial firms increase their cash holdings when monetary policy is loose. 28

29 To further test those two theories, we also include those firm size and policy deviation interaction variables as discussed in the last section in Models (2), (4), (6) and (8) of Panel A and B in Table 5 to see the behavioral difference between large and small firms. Take Model (2) from Panel B of Table 5 for example, when monetary policy is tight, the marginal effects of policy deviation is ( *Policy Deviation) for small firms. Therefore, the reaction of corporate cash holdings to policy deviations for small firms is a concave function, meaning when monetary policy is tight, small firms tend to increase their cash holdings with worsened asymmetric information problem, but when monetary policy is tighter, marginal increase of cash holdings will decrease. In comparison, when monetary policy is tight, the marginal effects of policy deviation is ( *Policy Deviation), which is a convex function. Therefore, when monetary policy is tightened, large industrial firms tend to reduce their cash holdings because they are less financially constrained, although when monetary policy is extremely tight (the policy deviation is greater than 8%, which does not exist in our research sample), large firms do increase their cash holdings. The log net cash ratio is more sensitive to the lagged policy deviation than the cash to total assets ratio with greater coefficients. Overall we find that there is a statistically significant positive relation between different measures of cash holdings and the lagged policy deviation. Our findings provide evidence supporting the credit channel prediction in the short run that industrial firms increase their cash holdings in response to monetary policy tightness, suggesting that industrial firms resort to their internal capital as a buffer for higher external funds premium or bondholders extra lending requirements caused by positive monetary shocks. 29

30 Also after comparing the corporate cash holdings changes between large and small firms when monetary policy is tight, we find that small firms increase their cash holdings due to tightened external financing environment while large firms reduce their cash holdings because they are less financial constrained and the Tobin s Q impacts dominate over the credit channel impacts. We also examine the impact of current monetary shocks, or the current policy deviation, on the corporate cash holdings and do not find consistent evidence supporting a possible relation for both annual and quarterly regression results (see Table 1 of the Appendix). To see whether these findings are caused by the possible endogeneity problem, we include four recent serial policy deviation variables to explain the corporate cash holdings in Table 2 of the Appendix. Both panels of Table 2 in the Appendix report evidence supporting a significantly positive relation between corporate cash holdings and the lagged policy deviation. Although quarterly regression results provide consistent evidence supporting a significantly negative relation between corporate cash holdings and current policy deviation from Panel B of Table 2 of the Appendix, annual regression results do not provide the same consistent evidence. To test whether keeping policy-controlled interest rates too low for too long inadvertently exacerbate financial imbalances through corporate cash holdings to buffer the monetary policy effectiveness, we examine the relation between corporate cash holdings and the cumulative policy deviation within the most recent four periods. Table 6 reports cash regressions results analogous to those in Table 5, except that we substitute cumulative policy deviation and its interaction variables for those policy deviation variables. 30

31 Yearly regression results of Models (1), (3), (5) and (7) from Panel A in Table 6 show that cumulative policy deviation is significantly positive for all those different regression models, for different measures of cash holdings and for different types of Taylor rule specification variables. Take Model (1) for example, a 1% continuous positive monetary shock leads to 5 basis points increase of corporate cash holdings in the U.S., keeping all else constant. These findings support the credit channel in the long run: when industrial firms face tight monetary policy within the recent four periods, industrial firms increase their cash holdings. Is it possible that the significance of the cumulative policy deviations is caused by any special previous policy deviations, but not by the production of new information? Table 2 in the Appendix already answers this question: only the lagged policy deviation is consistently and significantly positive for all firm fixed effects models for different types of Taylor rule prescriptions, different cash holdings measures and different frequency of our data sample. 2000s is special in that 2002 to 2006 saw the actual federal funds rate is below the values implied by the Taylor rule-by about 200 basis points on average over this fiveyear period (Taylor, 2007; Bernanke, 2010). As reported from Table 4, industrial firms also increase their cash holdings when monetary policy is extremely loose. We include the interaction variable Cumulative policy deviation 2000s dummy in Models (2), (4), (6) and (8) from Panel A in Table 6 to examine whether keeping policy-controlled interest rates too low for too long (Kahn, 2010) contribute to an increase in cash holdings in the 2000s that cannot be explained by changes in firm characteristics (Bates et al., 2009). It is also important to show the behavioral difference between large and 31

32 small firms, therefore we also include interaction variables Cumulative policy deviation large and Cumulative policy deviation large 2000s dummy. The inclusion of those interaction variables help decompose the impact of monetary shocks on corporate cash holdings into the pre-2000s and 2000s periods and decompose the impact of monetary shocks on small and large firms for those sub-periods. Take Model (2) from Panel B in Table 5.6 for example, for the pre-2000s period small firms have a coefficient of corporate cash holdings to monetary shocks of 0.089, implying a 1% continuous positive monetary shock leads to about 9 basis points corporate cash holdings increase for small firms. For the same period previous to the 2000s, large firms have a coefficient of corporate cash holdings to monetary shocks of , implying a 1% continuous positive monetary shock leads to about 3 basis points corporate cash holdings reduction for large firms. These findings are consistent with those of Table 5.5 in reporting that when monetary policy is (continuously) tight, small firms increase their corporate cash holdings while large firms reduce their cash holdings. For the same model, we find that for the 2000s period, small firms have a coefficient of corporate cash holdings to monetary shocks of , implying that for the 2000s period a 1% continuous negative monetary shock leads to around 2 basis points cash holdings increase for small firms, keeping all else constant i.e. small firms take advantage of the long lasting lower interest rate period to accumulate their cash holdings. Large firms have a coefficient of corporate cash holdings to monetary shocks of , implying that for the 2000s period a 1% continuous negative monetary shock leads to around 13 basis points cash holdings increase for large firms, keeping all else constant. This inverse relation between corporate cash holdings and cumulative policy 32

33 deviations we find in Table 6 shows that when the federal funds rate was set too long for too long, the Tobin s Q impact dominated the credit channel impact on corporate cash holdings i.e. industrial firms timed the general loose monetary environment to issue more stocks and the cumulative impact is a jump in their cash holdings compared with those in the 1980s and 1990s. Furthermore, large firms increased more their cash holdings than small ones because they are better at controlling their leverage. Overall our findings from Table 6 support the credit channel in the long run for the pre-2000s: industrial firms tend to increase their cash holdings when facing persistent tight monetary shocks in the long run. Because large firms are less financially constrained, the Tobin s Q effects dominate the credit channel effects for those firms, like what happened in the short run, large firms also reduce their cash holdings in response to the persistently tight monetary shocks. The 2000s period is special for its long lasting lower interest rate monetary environment, when the Federal Reserve sets its funds rate far below the Taylor rule prescription for so long. The cumulative effect as discussed in Baker and Wurgler (2002) is that industrial firms take the continuous loose monetary environment to issue more stocks when overpriced leading to jumped cash holdings. We find that both small and large firms increased their cash holdings in the 2000s in response to the continuous loosened monetary policy environment. Large firms are better managed at controlling their operation cost therefore are more reactive and increased their cash holdings more than those of small firms. We test the robustness of our conclusion regarding the impact of cumulative policy deviation on corporate cash holdings with two other different measures of cumulative policy deviation: cumulative policy deviation within the most recent eight 33

34 periods and twelve periods. Our results are robust for different measures of cumulative policy deviation, for different frequency of data sample and for different specifications of Taylor rule prescriptions (see Table 3 and Table 4 in the Appendix). We also examine how industrial firms change their cash holdings in response to the Taylor rule prescriptions, for Taylor rule is well-known and acknowledged by the Federal Reserve that the FOMC make monetary policy on this basis although not alone (Bernanke, 2010). Table 7 reports regressions of cash holdings on two different types of Taylor rule prescriptions and controls. We do not find consistent evidence supporting a significant relation between Taylor rule prescriptions and corporate cash holdings for either yearly data sample analysis or quarterly data sample analysis. As shown from Panel A in Table 7, firm fixed effects models (5) and (6) with cash ratio as the dependent variable show that there is a possible negative relation between Type 1 Taylor rule prescription and corporate cash holdings, but the relation is neither statistically nor economically significant when using Type 2 Taylor rule prescription. On the contrary, firm fixed effects models (7) and (8) provide evidence supporting a statistically significant positive relation between two types of Taylor rule prescriptions and corporate cash holdings when cash holdings is measured by the log value of net cash ratio. Although we find that the interaction variable Taylor prescription large is all negative across all those eight regression models, implying that large firms are less sensitive to Taylor rule prescriptions than small firms, the evidence is not all significant. We cannot find consistent evidence supporting a possible relation between corporate cash holdings and Taylor rule prescriptions either, when using quarterly data sample from Panel B in Table 7. 34

35 These findings regarding Taylor rule prescriptions is consistent with economic intuition that Taylor rule prescriptions provide essentially information about current inflation and output gap, but no information about firm income level or changes of shortterm real interest rates. Another possible explanation is that simple rules necessarily leave out many factors that may be relevant to the making of effective policy in a given episode (Bernanke, 2010) and industrial firms do not include them into their corporate decision making. Furthermore, there are no specific numerical values for those coefficients in Equation (1), and Taylor rule prescriptions may also depend sensitively on how inflation and the output gap are measured (Bernanke, 2010). Could federal funds rates and changes of federal funds rates be better monetary policy proxies? As discuss from the above, based on all those theories it is the monetary policy deviation that matters. To convince this point, we also report in the Appendix using either federal funds rates or changes in federal funds rates alone together with other firm specific and macro control variables to explain the formation of cash holdings. As reported from Table 7 of the Appendix, neither federal funds rates nor changes in federal funds rates are consistently significant in explaining the formation of cash holdings for different types of Taylor rule specifications, for different models, for different cash ratio measures and for different frequency of data samples. We test three theories regarding the possible relation between corporate cash holdings and monetary shocks: interest rate channel, Tobin s Q and credit channel. Interest rate channel emphasizes the opportunity cost of holding cash, although the prediction is that when interest rate is higher, industrial firms tend to hold less cash, our findings regarding the relation between corporate cash holdings and the Funds rate is 35

36 consistent with Bates et al. (2009) in stating a non-significant relation between corporate cash holdings and the Funds rates because firms and financial intermediaries have become more efficient in handling transactions. Small firms are more likely to be financially constrained, which makes the asymmetric information problem when getting external funding. Therefore, the relation between corporate cash holdings and monetary shocks for small firms is more supported by the credit channel. We find that for small firms, both in the short-term or in the long run (for the pre-2000s period), there is a significant positive relation between corporate cash holdings and transient (and cumulative) monetary shocks, supporting the credit channel prediction. Because small firms are more financially constrained, liquidity problem becomes much severer when monetary policy is tighter, and it becomes much harder to squeeze cash out of their financial statements. To be more specific, we find that the relation between corporate cash holdings and transient monetary shocks is a concave function, implying that small industrial firms increase their cash holdings when experiencing tight monetary shocks and the marginal increase of cash holdings decreases with the further tightening of monetary policy. On the other hand, large firms are said to be less or even not financially constrained, therefore the asymmetric information problem becomes not that important. The benefits of reducing cash holdings dominates the costs of increasing cash holdings, therefore large firms have more flexibility to weight their relative benefits and costs. Because large firms are better at managing their stock values and setting their capital structure, industrial firms are more likely to take advantage of changing monetary looseness and tightness to increase their stock issuance when their stocks are over-valued 36

37 and to increase their stock repurchases when their stocks are under-valued. Overall we find an inverse relation between corporate cash holdings and monetary shocks, both in the short-run and in the long-run (for the pre-2000s period), implying that when monetary policy is tight, large industrial firms reduce their cash holdings. The short-term relation between corporate cash holdings and monetary shocks is a convex function, so when monetary policy is extremely tight (which means when the policy deviation is wider than 8%), large industrial firms also increase their cash holdings because the impact of credit channel (problem of asymmetric information) dominates the impact of Tobin s Q. Also we find that large firms reduce their cash holdings when monetary policy is continuously tight, also supporting the Tobin s Q theory in stating that industrial firms increase their stock repurchases when stocks are undervalued, leading to reduced cash holdings. We also find that both large and small firms take the opportunity of loose monetary environment in the 2000s when stocks were more likely to be overpriced, leading to more stock issuance and therefore increased corporate cash holdings. Conclusion This research examines how the effect of monetary policy tightness or ease on corporate cash holdings to better understand how monetary policy targeting at price stability and unemployment, economic growth influence current-future investment conflicts. We do not find significant evidence regarding the relation between corporate cash holdings and the Funds rate, not supporting the interest rate channel prediction about an inverse relation between those two variables. Small firms are less financially constrained therefore the credit channel effects dominate the Tobin s Q effects. We find a positive relation between corporate cash holdings and transient policy deviation for small 37

38 firms in the short-run and this positive relation extends to the long-run for the pre-2000s period. Large firms are less financially constrained therefore the Tobin s Q effects dominate the credit channel effects. We find a negative relation between corporate cash holdings and transient monetary shocks for large firms in the short-run and this relation extends to the long run for the pre-2000s period. The 2000s is specified by [the Fed] keeping policy-controlled interest rates too low for too long (Kahn, 2010), when monetary policy was too easy during the period from 2002 to 2006, as the actual federal funds rate is below the values implied by the Taylor rule-by about 200 basis points on average over this five-year period (Taylor, 2007; Bernanke, 2010). Our findings provide empirical evidence about an inverse relationship between sustained policy deviation and corporate cash holdings in the 2000s for both large and small industrial firms in the U.S. Large firms are better managed at controlling their operation cost therefore are more reactive and increased their cash holdings more than those of small firms. The evidence suggests that policymakers should monitor financial conditions for signs that cash are hoarding for industrial firms. Although policymakers may have many reasons to deviate from simple rule-like behavior, they should be alert to unintended consequences from maintaining rates too low for too long. Our findings raise serious concerns about the current practice when too much cash becomes a really serious business problem 20. Our study urges more exploration on this topic in the future

39 39 Table 1 Theories and Predictions Regarding the Relation between Monetary Policy Variables and Cash Holdings Transmission mechanisms of monetary policy Interest rate channel Tobin s q theory Credit channel Theory Predicted relationship Policy deviation Cumulative policy deviation Expansionary monetary A negative relationship (-) (-) policy leads to a fall in real between monetary policy No prediction about the monetary policy interest rates, which in turn tightness and corporate cash impact difference on small and large firms lowers the opportunity cost holdings. of holding cash. Expansionary monetary policy leads to a rise in stock prices, industrial firms then take the opportunity to issue more equities. Corporate cash holdings increase following equity issues. Expansionary monetary policy helps reduce the external finance premium, and increase significantly the rate of inflation, resulting in decreasing cash holding. A negative relationship between monetary policy tightness and corporate cash holdings in the long run. A positive relationship between monetary policy tightness and corporate cash holdings. Monetary policy will have a greater effect on smaller firms that are more dependent on bank loans than it will on large firms that can directly access the credit markets through other markets. (-) (-) No prediction about the monetary policy impact difference on small and large firms (+) (+) Monetary policy impact more on small firms than on large firms

40 Table 2 Taylor Rule Parameterizations (Kahn, 2010) r α β Rule Rule Rule Rule Note: Taylor rule prescriptions prescribe the Federal Reserve should follow in setting the federal funds rate in the general Taylor rule form: i t = r + π t + α(π t π t ) + β(y t y t ), where i t represents the recommended policy rate as measured by the federal funds rate, r represents the equilibrium real interest rate, (π t π t ) represents the deviation of the inflation rate (π t ) from its long-run target (π t ), (y t y t ) represents the output gap the level of real GDP (y t ) relative to potential GDP (y t ), and α, β are parameters. This table identifies the four specifications of the Taylor rule to be examined. All of the rules adhere to the Taylor principle that policymakers should adjust the nominal federal funds rate more than one-for-one with an increase in inflation relative to target. Rule 1 is the original version of the Taylor rule (Taylor, 1993). Rule 2 places a higher weight on output than the original Taylor rule (Ball, 1997). Rule 3 and Rule 4 assume higher equilibrium real rates and different weights on inflation and the output gap. For the calculation, we get real GDP from Bureau of Economic Analysis, potential real GDP from Congressional Budget Office, and CPI data from Bureau of Labor Statistics. Inflation is measured by the four-quarter rate of change in the CPI and the output gap measured as the log ratio of real GDP to the CBO estimate of potential. 40

41 Table 3 Descriptive Statistics and Correlations of Control Variable Characteristics and Monetary Policy Variables Panel A-1: Yearly descriptive statistics Lower Upper Std Dev Variable Mean Quartile Median Quartile Minimum Maximum N Cash/Assets Taylor prescription Policy deviation Squared policy deviation Cumulative policy deviation Taylor prescription Policy deviation Squared policy deviation Cumulative policy deviation Fiscal deficit Credit spread Industry sigma Market to book Real size Cash flow/assets NWC/assets Capex Leverage R&D/sales Dividend dummy Acquisition activity Net debt issuance Net equity issuance Loss dummy Note: The yearly data sample includes all Compustat firm-year observations from 1980 to 2007 with positive values for the book value of total assets and sales revenue for firms incorporated in the United States. Financial firms (SIC code ) and utilities (SIC codes ) are excluded from the sample, yielding a panel of 118,897 firm-year observations for 13,743 unique firms. Missing explanatory values reduce the panel used here to 77,738 firm-year observations for 12,430 unique firms for the OLS regressions. The quarterly sample yields a panel of 439,659 firm-quarter observations for 13,210 unique firms. Missing explanatory values reduce the panel used here to 218,502 firm-quarter observations for 10,636 unique firms for the OLS regressions. Panel A reports descriptive statistics, and Panel B reports Pearson correlation coefficients together with p-values for the significance. Cash/Assets is cash and marketable securities (data item #1) divided by total assets (data item #6). Taylor prescriptions are calculated based on two types of Taylor rule specifications from Table 2. Policy deviation is the difference between the actual annual average federal funds rates and Taylor prescriptions. Squared policy deviation is the squared value of policy deviation. Cumulative policy deviations are the sum of Taylor rule deviations from the first period up to the current. We calculate the fiscal deficit as the difference between annual federal government current receipts and current expenditures divided by nominal GDP. Credit spread is the difference between the AAA and BBB yields reported by the Federal Reserve. Industry sigma is the average across the two-digit SIC code of the firm cash flow standard deviations for the previous 10 years, and we require at least three observations for the calculation. Market-tobook is the ratio of the market value of assets to the book value of assets i.e. book value of assets (#6) minus the book value of equity (#60) plus the market value of equity (#199* #25) as the numerator of the ratio and the book value of assets (#6) as the denominator. Real size is the logarithm of book assets (#6). Cash flow/assets is calculated as earnings after interest, dividends, and taxes but before depreciation divided by book assets (((#13 #15 #16 #21)/#6). NWC/assets is net working capital (data item #179) minus cash and marketable securities (data item #1) divided by book assets. Capex is the ratio of capital expenditures (data item #128) to the book value of total assets (data item #6). Leverage is the ratio of total debt to the book value of total assets (data item #6), where debt includes long-term debt (data item #9) plus debt in current liabilities (data item #34). R&D/sales is the ratio of research and development expense (data item #46) to sales (data item #12). Dividend dummy is a dummy variable equal to one if the firm paid a common dividend and zero otherwise. Acquisition activity is the ratio of expenditures on acquisitions (data item #129) relative to the book value of total assets (data item #6). Net debt issuance is calculated as annual total debt issuance (data item #111) minus debt retirement (data item #114), divided by the book value of total assets (data item #6). Net equity issuance is calculated as equity sales (data item #108) minus equity purchases (data item #115), divided by the book value of total assets (data item #6). Loss dummy is a dummy variable equal to one if net income (data item #172) is less than zero, and zero otherwise. All variables in dollars are inflationadjusted to 2007 dollars using the Consumer Price Index. 41

42 Table 3 (continued) Panel A-2: Quarterly descriptive statistics Variable Mean Lower Upper Median Quartile Quartile Std Dev Minimum Maximum N Cash/Assets Taylor prescription Policy deviation Squared policy deviation Cumulative policy deviation Taylor prescription Policy deviation Squared policy deviation Cumulative policy deviation Fiscal deficit Credit spread Industry sigma Market to book Real size Cash flow/assets NWC/assets Capex Leverage R&D/sales Dividend dummy Acquisition activity Net debt issuance Net equity issuance Loss dummy

43 43 Table 3 (continued) Panel B-1: Yearly correlations between cash, monetary policy variables, and macro control variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) 1 Cash/Assets Taylor prescription < Policy deviation <.0001 < Squared policy deviation Cumulative policy deviation 1 <.0001 <.0001 < <.0001 <.0001 <.0001 < Fiscal deficit <.0001 <.0001 <.0001 <.0001 < Credit spread <.0001 <.0001 <.0001 <.0001 <.0001 < Taylor prescription <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 < Policy deviation <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 < Squared policy deviation 2 11 Cumulative policy deviation <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 < <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001

44 44 Table 3 (continued) Panel B-2: Quarterly correlations between cash, monetary policy variables, and macro control variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) 1 Cash/Assets Taylor prescription < Policy deviation <.0001 < Squared policy deviation <.0001 <.0001 < Cumulative policy deviation <.0001 <.0001 <.0001 < Fiscal deficit <.0001 <.0001 <.0001 <.0001 < Credit spread <.0001 <.0001 <.0001 <.0001 < Taylor prescription <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 < Policy deviation <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 < Squared policy deviation <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 < Cumulative policy deviation <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001

45 45 Changes of Firm Characteristics by Policy Deviation Quartiles Panel A-1: Yearly monetary policy regimes specified by Type 1 policy deviations Loose monetary policy regime Tight monetary policy regime First quartile Second quartile Third quartile Fourth quartile First quartile Second quartile Third quartile Fourth quartile Policy deviation interval [-0.064, ] (-0.024, ] (-0.013, ] (-0.008, ] [0.000, 0.006] (0.006, 0.013] (0.013, 0.030] (0.030, 0.061] ΔCash/Assets [0.001] [0.000] [-0.000] [-0.001] [-0.001] [-0.001] [-0.001] [0.001] Δcash [0.345] [0.040] [-0.008] [0.000] [0.000] [-0.019] [-0.028] [0.035] Δindustry sigma [0.000] [0.000] [-0.002] [0.000] [0.000] [0.001] [0.002] [0.003] Δmarket to book [0.063] [0.017] [-0.035] [-0.064] [0.032] [0.001] [0.000] [0.023] Δreal size [0.016] [0.000] [0.003] [0.025] [0.015] [0.026] [0.020] [0.013] [-0.001] [-0.001] [0.000] [0.001] [0.001] [0.000] [-0.003] [0.000] Δcash flow/assets ΔNWC/assets [0.000] [-0.004] [-0.001] [-0.001] [-0.002] [-0.001] [-0.005] [-0.005] Δcapex [0.001] [-0.002] [-0.001] [-0.001] [-0.001] [0.000] [0.000] [-0.001] Δleverage [-0.003] [-0.002] [-0.001] [0.000] [-0.001] [0.000] [0.000] [-0.004] ΔR&D/sales [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δdividend dummy [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]

46 46 Table 4 (continued) Δacquisition activity [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δnet debt issuance [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δnet equity issuance [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δloss dummy [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Panel A-2: Quarterly monetary policy regimes specified by Type 1 policy deviations Policy deviation interval ΔCash/Assets Δcash Δindustry sigma Δmarket to book Δreal size Δcash flow/assets ΔNWC/assets Δcapex Loose monetary policy regime Tight monetary policy regime First Second Third Fourth First Second Third Fourth quartile quartile quartile quartile quartile quartile quartile quartile [0.000] [0.000] [0.000] [-0.001] [0.000] [-0.001] [0.000] [-0.001] [0.046] [0.000] [-0.038] [-0.056] [0.000] [-0.070] [0.000] [-0.055] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.016] [0.013] [-0.020] [-0.014] [-0.001] [0.015] [-0.018] [-0.002] [0.004] [-0.004] [-0.001] [0.005] [0.003] [0.006] [0.010] [0.002] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.001] [-0.001] [0.000] [0.001] [-0.001] [0.002] [-0.001] [0.000] [0.004] [0.006] [0.006] [0.007] [0.008] [0.005] [0.008] [0.009]

47 47 Table 4 (continued) Δleverage [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] ΔR&D/sales [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δdividend dummy [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δacquisition activity [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δnet debt issuance [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δnet equity issuance [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δloss dummy [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Panel B-1: Yearly monetary policy regimes specified by Type 2 policy deviations Loose monetary policy regime Tight monetary policy regime First quartile Second quartile Third quartile Fourth quartile First quartile Second quartile Third quartile Fourth quartile Policy deviation interval [-0.059, ] (-0.028, ] (-0.018, ] (-0.007, ] [0.000, 0.004] (0.004, 0.012] (0.012, 0.028] (0.028, 0.094] ΔCash/Assets [0.000] [0.001] [0.000] [-0.001] [-0.001] [-0.001] [0.000] [0.001] Δcash [0.048] [0.299] [0.020] [-0.027] [-0.018] [0.004] [-0.007] [0.033] Δindustry sigma [0.001] [0.000] [-0.002] [0.000] [0.000] [0.001] [0.001] [0.003] Δmarket to book [0.003] [0.059] [-0.054] [-0.012] [0.009] [-0.035] [0.036] [0.021]

48 48 Table 4 (continued) Δreal size [0.008] [0.016] [0.003] [0.008] [0.024] [0.039] [0.008] [0.011] Δcash flow/assets [-0.002] [-0.002] [-0.002] [0.001] [0.002] [0.001] [-0.002] [-0.001] ΔNWC/assets [0.000] [-0.004] [-0.003] [-0.001] [0.000] [-0.002] [-0.004] [-0.006] Δcapex [0.001] [-0.002] [-0.001] [0.000] [0.001] [0.000] [-0.002] [-0.001] Δleverage [-0.002] [0.000] [-0.001] [-0.003] [0.000] [0.000] [-0.001] [-0.004] ΔR&D/sales [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δdividend dummy [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δacquisition activity [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δnet debt issuance [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δnet equity issuance [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Δloss dummy [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Panel B-2: Quarterly monetary policy regimes specified by Type 2 policy deviations Loose monetary policy regime Tight monetary policy regime First quartile Second quartile Third quartile Fourth quartile First quartile Second quartile Third quartile Fourth quartile Policy deviation interval [-0.038, ] (-0.025, ] (-0.017, ] (-0.007, ] [0.000, 0.004] (0.004, 0.010] (0.010, 0.022] (0.022, 0.096]

49 49 Table 4 (continued) ΔCash/Assets Δcash Δindustry sigma Δmarket to book Δreal size Δcash flow/assets ΔNWC/assets Δcapex Δleverage ΔR&D/sales Δdividend dummy Δacquisition activity Δnet debt issuance Δnet equity issuance [0.000] [0.000] [0.000] [-0.001] [-0.001] [0.000] [-0.001] [-0.001] [0.000] [0.013] [-0.005] [-0.035] [-0.038] [-0.018] [-0.020] [-0.036] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.005] [-0.003] [-0.008] [-0.002] [-0.003] [-0.018] [0.013] [0.000] [0.003] [0.002] [-0.006] [0.002] [0.007] [0.005] [0.009] [0.001] [0.000] [0.000] [0.000] [0.000] [0.000] [0.001] [0.000] [0.000] [0.001] [0.000] [-0.001] [0.001] [0.000] [0.000] [0.001] [0.000] [0.005] [0.005] [0.006] [0.006] [0.008] [0.008] [0.004] [0.009] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]

50 50 Table 4 (continued) Δloss dummy [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] Note: Univariate comparison of means and medians of measures of firm characteristics changes of U.S.-based publicly traded firms. Median values are bracketed. Cash is cash and marketable securities (data item #1). Cash/Assets is cash and marketable securities (data item #1) divided by total assets (data item #6). Taylor prescriptions are calculated based on two types of Taylor rule specifications from Table 2. Policy deviation is the difference between the actual annual average federal funds rates and Taylor prescriptions. Squared policy deviation is the squared value of policy deviation. Cumulative policy deviations are the sum of Taylor rule deviations from the first period up to the current. We calculate the fiscal deficit as the difference between annual federal government current receipts and current expenditures divided by nominal GDP. Credit spread is the difference between the AAA and BBB yields reported by the Federal Reserve. Industry sigma is the average across the two-digit SIC code of the firm cash flow standard deviations for the previous 10 years, and we require at least three observations for the calculation. Market-to-book is the ratio of the market value of assets to the book value of assets i.e. book value of assets (#6) minus the book value of equity (#60) plus the market value of equity (#199* #25) as the numerator of the ratio and the book value of assets (#6) as the denominator. Real size is the logarithm of book assets (#6). Cash flow/assets is calculated as earnings after interest, dividends, and taxes but before depreciation divided by book assets (((#13 #15 #16 #21)/#6). NWC/assets is net working capital (data item #179) minus cash and marketable securities (data item #1) divided by book assets. Capex is the ratio of capital expenditures (data item #128) to the book value of total assets (data item #6). Leverage is the ratio of total debt to the book value of total assets (data item #6), where debt includes long-term debt (data item #9) plus debt in current liabilities (data item #34). R&D/sales is the ratio of research and development expense (data item #46) to sales (data item #12). Dividend dummy is a dummy variable equal to one if the firm paid a common dividend and zero otherwise. Acquisition activity is the ratio of expenditures on acquisitions (data item #129) relative to the book value of total assets (data item #6). Net debt issuance is calculated as annual total debt issuance (data item #111) minus debt retirement (data item #114), divided by the book value of total assets (data item #6). Net equity issuance is calculated as equity sales (data item #108) minus equity purchases (data item #115), divided by the book value of total assets (data item #6). Loss dummy is a dummy variable equal to one if net income (data item #172) is less than zero, and zero otherwise. All variables in dollars are inflation-adjusted to 2007 dollars using the Consumer Price Index. ΔXt is notation for the one-year change, Xt- Xt-1, where t and (t-1) denote end of fiscal year t and (t-1). When monetary policy deviation, the actual federal funds rate minus the Type 1 (or 2) Taylor rule prescription, is positive, the economy is defined as in tight monetary policy regime ; when monetary policy deviation is negative, the economy is defined as in loose monetary policy regime. In this analysis, we first divide the whole sample into two subsamples, one subsample with positive policy deviations implying too tight monetary policy; the other subsample with negative policy deviations implying too loose monetary policy. Within each subsample we divide them into four quartiles based on the value of negative or positive policy deviations.

51 51 Regressions of Cash Holdings on Policy Deviation Variables and Controls Panel A: Yearly regression results Types of Taylor rule T. 1 T. 2 T. 1 T. 2 Dependent variable Cash/Assets Log(Cash/Net Assets) Independent variable (1) (2) (3) (4) (5) (6) (7) (8) Policy deviation t *** *** 0.074*** *** ** * (5.64) (4.22) (3.87) (3.34) (1.94) (2.90) (-1.51) (2.06) (Policy Dev. t-1 ) *** *** ** 47.10*** *** 21.06*** *** (3.53) (4.58) (1.08) (2.99) (8.17) (5.67) (6.46) (4.73) (Policy Dev. t-1 ) 2 tight ** * * ** (-2.84) (-2.39) (-2.48) (-2.97) Policy deviation t tight large (-1.10) (-1.22) (-1.27) (-1.51) (Policy Dev. t-1 ) 2 tight large (-0.13) (0.09) (0.78) (1.07) Funds rate (1.19) (0.66) (1.68) (1.52) (0.46) (-0.02) (1.36) (1.12) Fiscal deficit *** *** *** *** *** *** *** *** (-3.70) (-4.01) (-3.37) (-3.32) (-5.98) (-6.13) (-6.49) (-6.42) Credit spread 1.569*** *** 1.629*** *** 27.91*** *** 27.87*** *** (7.17) (6.84) (7.50) (7.37) (11.34) (10.99) (11.40) (11.21) Industry sigma * * * * 1.942*** *** 1.861*** *** (2.13) (2.30) (2.08) (2.31) (4.58) (4.70) (4.41) (4.63) Market to book 0.007*** *** 0.007*** *** *** *** *** *** (7.72) (7.61) (7.71) (7.62) (10.53) (10.43) (10.58) (10.49) Real size ** * ** * (-0.52) (-0.29) (-0.51) (-0.31) (-2.72) (-2.49) (-2.69) (-2.42) Cash flow/assets * * 0.162* * (1.65) (1.61) (1.63) (1.60) (2.34) (2.30) (2.33) (2.29)

52 52 Table 5 (continued) NWC/assets *** *** *** *** *** *** *** *** (-34.47) (-34.50) (-34.46) (-34.47) (-30.76) (-30.78) (-30.72) (-30.75) Capex *** *** *** *** *** *** *** *** (-27.13) (-27.19) (-27.09) (-27.15) (-20.87) (-20.93) (-20.78) (-20.89) Leverage *** *** *** *** *** *** *** *** (-33.34) (-33.39) (-33.35) (-33.38) (-36.36) (-36.39) (-36.39) (-36.44) R&D/sales 0.016*** *** *** *** *** *** 0.099*** *** (3.50) (3.51) (3.50) (3.51) (3.74) (3.76) (3.75) (3.77) Dividend dummy (-0.18) (-0.16) (-0.16) (-0.15) (-0.91) (-0.90) (-0.88) (-0.89) Acquisition activity *** *** *** *** *** *** *** *** (-26.00) (-26.01) (-25.94) (-25.95) (-16.64) (-16.67) (-16.50) (-16.54) Net debt issuance 0.159*** *** 0.159*** *** 1.188*** *** 1.190*** *** (15.37) (15.37) (15.37) (15.37) (12.53) (12.52) (12.53) (12.53) Net equity issuance 0.174*** *** 0.174*** *** 1.141*** *** 1.140*** *** (23.02) (22.98) (23.02) (22.99) (18.96) (18.90) (18.93) (18.88) Loss dummy *** *** *** *** *** *** *** *** (-17.64) (-17.61) (-17.66) (-17.63) (-14.70) (-14.69) (-14.75) (-14.72) Intercept 0.223*** *** 0.223*** *** *** *** *** *** (21.65) (21.55) (21.55) (21.23) (-17.69) (-17.74) (-17.81) (-17.84) Within R Panel B: Quarterly regression results Types of Taylor rule T. 1 T. 2 T. 1 T. 2 Dependent variable Cash/Assets Log(Cash/Net Assets) Independent variable (1) (2) (3) (4) (5) (6) (7) (8) Policy deviation t *** ** *** ** 1.703** (1.03) (5.06) (-0.04) (3.25) (-1.69) (6.65) (-2.83) (2.73)

53 53 Table 5 (continued) (Policy Dev. t-1 ) *** 15.44*** 1.915** 8.059*** 78.60*** 294.3*** 29.42*** 127.1*** (5.00) (6.01) (3.19) (3.97) (7.56) (10.16) (4.36) (5.46) (Policy Dev. t-1 ) 2 tight *** ** *** *** (-4.48) (-3.00) (-8.52) (-4.54) Policy deviation t *** *** *** *** tight large (-4.44) (-4.97) (-4.31) (-4.44) (Policy Dev. t-1 ) 2 tight * 116.9** 71.78** large (1.51) (2.09) (3.10) (3.20) Funds rate 0.163*** 0.160*** 0.171*** 0.167*** (4.66) (4.54) (4.92) (4.77) (1.80) (1.51) (1.69) (1.42) Fiscal deficit *** *** *** *** *** *** *** *** (-4.99) (-4.67) (-5.56) (-5.54) (-5.91) (-5.29) (-7.35) (-7.25) Credit spread 1.774*** 1.935*** 1.877*** 1.992*** 27.95*** 31.00*** 29.45*** 31.17*** (7.07) (7.53) (7.42) (7.66) (9.78) (10.58) (10.23) (10.52) Industry sigma * * (-1.88) (-1.87) (-1.98) (-2.02) (-1.57) (-1.50) (-1.82) (-1.87) Market to book 0.007*** 0.007*** 0.007*** 0.007*** *** 0.081*** *** 0.082*** (9.95) (9.84) (10.01) (9.87) (13.13) (12.98) (13.25) (13.07) Real size * (0.38) (0.38) (0.49) (0.62) (-1.92) (-2.09) (-1.72) (-1.62) Cash flow/assets *** *** *** *** (-3.36) (-3.37) (-3.37) (-3.40) (-0.62) (-0.62) (-0.65) (-0.67) NWC/assets *** *** *** *** *** *** *** *** Capex (-30.27) (-30.28) (-30.25) (-30.27) (-28.72) (-28.70) (-28.73) (-28.71) *** *** *** *** *** *** *** *** (-20.07) (-20.09) (-20.02) (-20.09) (-9.83) (-9.86) (-9.87) (-9.96) Leverage *** *** *** *** *** *** *** *** (-31.83) (-31.83) (-31.90) (-31.92) (-35.83) (-35.80) (-35.95) (-35.96)

54 54 Table 5 (continued) R&D/sales 0.009*** 0.009*** 0.009*** 0.009*** *** 0.058*** *** 0.058*** (3.82) (3.85) (3.82) (3.85) (4.08) (4.10) (4.08) (4.12) Dividend dummy * * * * * * * * (-2.17) (-2.00) (-2.16) (-1.96) (-2.50) (-2.40) (-2.48) (-2.35) Acquisition activity *** *** *** *** *** *** *** *** (-23.20) (-23.26) (-23.11) (-23.11) (-11.26) (-11.42) (-11.18) (-11.19) Net debt issuance 0.160*** 0.160*** 0.161*** 0.161*** 1.091*** 1.089*** 1.099*** 1.100*** Net equity issuance (15.83) (15.83) (15.85) (15.86) (13.00) (12.99) (13.06) (13.07) 0.184*** 0.184*** 0.183*** 0.183*** 1.209*** 1.215*** 1.205*** 1.202*** (22.45) (22.46) (22.43) (22.41) (20.58) (20.64) (20.54) (20.51) Loss dummy *** *** *** *** *** *** *** *** (-11.22) (-11.25) (-11.24) (-11.25) (-12.00) (-12.03) (-12.02) (-12.04) Intercept 0.199*** 0.198*** 0.197*** 0.196*** *** *** *** *** (19.54) (19.43) (19.36) (19.17) (-18.90) (-18.96) (-19.14) (-19.31) Within R Note: Cash/Assets defined as cash and marketable securities (data item #1) divided by total assets (data item #6). Log net cash ratio defined as log value of cash and marketable securities (data item #1) divided by (total assets (data item #6)-cash and marketable securities (data item #1)). The sample includes all Compustat firm-year observations from 1980 to 2007 with positive values for the book value of total assets and sales revenue for firms incorporated in the United States. Financial firms (SIC code ) and utilities (SIC codes ) are excluded from the sample, yielding a panel of 118,897 observations for 13,743 unique firms. Missing explanatory values reduce the panel used here to 67,066 firm-year observations for 12,430 unique firms for the firm fixed effects regressions. Taylor prescriptions are calculated based on two types of Taylor rule specifications from Table 2. Policy deviation is the difference between the actual annual average federal funds rates and Taylor prescriptions. Squared policy deviation is the squared value of policy deviation. We calculate the fiscal deficit as the difference between annual federal government current receipts and current expenditures divided by nominal GDP. Credit spread is the difference between the AAA and BBB yields reported by the Federal Reserve. Industry sigma is the average across the two-digit SIC code of the firm cash flow standard deviations for the previous 10 years, and we require at least three observations for the calculation. Market-to-book is the ratio of the market value of assets to the book value of assets i.e. book value of assets (#6) minus the book value of equity (#60) plus the market value of equity (#199* #25) as the numerator of the ratio and the book value of assets (#6) as the denominator. Real size is the logarithm of book assets (#6). Cash flow/assets is calculated as earnings after interest, dividends, and taxes but before depreciation divided by book assets (((#13 #15 #16 #21)/#6). NWC/assets is net working capital (data item #179) minus cash and marketable securities (data item #1) divided by book assets. Capex is the ratio of capital expenditures (data item #128) to the book value of total assets (data item #6). Leverage is the ratio of total debt to the book value of total assets (data item #6), where debt includes long-term debt (data item #9) plus debt in current liabilities (data item #34). R&D/sales is the ratio of research and development expense (data item #46) to sales (data item #12). Dividend dummy is a dummy variable equal to one if the firm paid a common dividend and zero otherwise. Acquisition activity is the ratio of expenditures on acquisitions (data item #129) relative to the book value of total assets (data item #6). Net debt issuance is calculated as annual total debt issuance (data item #111) minus debt retirement (data item #114), divided by the book value of total assets (data item #6). Net equity issuance is calculated as equity sales (data item #108) minus equity purchases (data item #115), divided by the book value of total assets (data item #6). Loss dummy is a dummy variable equal to one if net income (data item #172) is less than zero, and zero otherwise. 2000s dummy is a dummy variable equal to one if the firm observation is in the fiscal year after 1999, and zero otherwise. All variables in dollars are inflation-adjusted to 2007 dollars using the Consumer Price Index. ΔX t is notation for the one-year change, X t- X t-1, where t and (t-1) denote end of fiscal year t and (t-1). In this analysis, we first divide the whole sample into four quartiles each fiscal year based on the real size and define firms in the largest real size quartiles as large firms. Large is a dummy variable equal to one if the firm is in the large real size quartile and zero otherwise. t-statistics based on standard errors robust to clustering by firm and year are reported in parentheses. We report adjusted-r 2 for OLS estimation models and within R 2 for firm fixed effects estimation models. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

55 55 Regressions of Cash Holdings on One Year Cumulative Policy Deviation Variables and Controls Panel A: Yearly regression results Types of Taylor rule T. 1 T. 2 T. 1 T. 2 Dependent variable Cash/Assets Log(Cash/Net Assets) Independent variable (1) (2) (3) (4) (5) (6) (7) (8) Cumulative policy deviation *** *** large 2000s dummy (0.32) (0.33) (-3.64) (-3.32) Cumulative policy deviation large (-0.97) (-1.67) (0.98) (0.33) Cumulative policy deviation *** *** *** *** 2000s dummy (-3.92) (-4.13) (-5.13) (-5.29) Cumulative policy deviation *** *** *** *** 0.212* *** 0.203* *** (6.47) (7.75) (6.50) (8.07) (2.31) (5.09) (2.49) (5.54) Funds rate * *** *** 0.931** *** 0.859* *** (2.24) (3.85) (1.71) (4.34) (2.60) (5.09) (2.45) (6.37) Fiscal deficit *** *** *** ** *** *** *** *** (-4.38) (-3.34) (-3.41) (-3.19) (-7.13) (-5.50) (-6.71) (-6.38) Credit spread 1.723*** *** 1.801*** *** 29.19*** *** 29.57*** *** (7.76) (7.33) (8.11) (7.11) (11.68) (11.05) (11.80) (10.25) Industry sigma *** ** 1.627*** ** (1.44) (0.78) (1.67) (0.99) (3.80) (2.77) (3.88) (2.86) Market to book 0.007*** *** 0.007*** *** 0.077*** *** 0.076*** *** (7.87) (7.72) (7.80) (7.75) (10.61) (10.29) (10.58) (10.42) Real size * *** * *** (-0.00) (-0.87) (-0.07) (-0.82) (-2.17) (-3.97) (-2.19) (-3.82) Cash flow/assets * ** 0.162* ** (1.65) (1.83) (1.67) (1.83) (2.32) (2.73) (2.33) (2.68) NWC/assets *** *** *** *** *** *** *** *** (-34.51) (-34.18) (-34.53) (-34.26) (-30.82) (-30.11) (-30.83) (-30.22)

56 56 Table 6 (continued) Capex *** *** *** *** *** *** *** *** (-27.14) (-26.72) (-27.18) (-26.84) (-20.88) (-20.08) (-20.92) (-20.21) Leverage *** *** *** *** *** *** *** *** (-33.57) (-33.20) (-33.56) (-33.27) (-36.60) (-36.02) (-36.62) (-36.16) R&D/sales *** *** *** *** 0.101*** *** 0.101*** *** (3.55) (3.48) (3.54) (3.49) (3.84) (3.74) (3.84) (3.77) Dividend dummy (-0.09) (-0.02) (-0.10) (0.00) (-0.76) (-0.67) (-0.76) (-0.62) Acquisition activity *** *** *** *** *** *** *** *** (-26.10) (-26.25) (-26.15) (-26.12) (-16.73) (-16.92) (-16.76) (-16.72) Net debt issuance 0.159*** *** 0.159*** *** 1.195*** *** 1.194*** *** (15.38) (15.43) (15.37) (15.40) (12.56) (12.64) (12.55) (12.60) Net equity issuance 0.175*** *** 0.175*** *** 1.147*** *** 1.147*** *** (23.07) (23.29) (23.08) (23.20) (19.06) (19.63) (19.06) (19.38) Loss dummy *** *** *** *** *** *** *** *** (-17.78) (-17.70) (-17.76) (-17.72) (-14.87) (-14.67) (-14.86) (-14.75) Intercept 0.219*** *** 0.220*** *** *** *** *** *** (21.15) (21.43) (21.18) (21.39) (-18.02) (-16.31) (-18.02) (-16.81) Within R Panel B: Quarterly regression results Types of Taylor rule T. 1 T. 2 T. 1 T. 2 Dependent variable Cash/Assets Log(Cash/Net Assets) Independent variable (1) (2) (3) (4) (5) (6) (7) (8) Cumulative policy deviation * large 2000s dummy Cumulative policy deviation large (1.36) (2.08) (-0.70) (0.96) *** *** ** *** (-4.82) (-5.55) (-2.64) (-3.52)

57 57 Table 6 (continued) Cumulative policy deviation *** * *** *** 2000s dummy (-3.74) (-2.45) (-6.11) (-3.73) Cumulative policy deviation *** *** *** * (1.75) (5.10) (1.18) (3.83) (-1.42) (4.43) (-1.88) (2.03) Funds rate 0.148*** *** 0.166*** 0.172*** 0.841* 0.953* 0.842* 0.881* (4.15) (4.41) (4.77) (4.91) (2.02) (2.31) (2.09) (2.18) Fiscal deficit *** *** *** *** *** *** *** *** (-5.81) (-6.02) (-5.10) (-5.28) (-7.06) (-7.33) (-7.15) (-7.34) Credit spread 1.871*** *** 1.932*** 1.822*** 31.29*** 23.31*** 31.01*** 28.58*** (6.93) (6.18) (7.29) (7.37) (10.14) (8.41) (10.25) (10.13) Industry sigma * * * * * (-2.13) (-1.88) (-2.08) (-2.02) (-1.97) (-1.40) (-2.01) (-1.87) Market to book 0.008*** *** 0.008*** 0.007*** 0.084*** *** *** *** (10.10) (9.85) (10.11) (9.89) (13.46) (12.98) (13.47) (13.16) Real size ** (0.70) (0.19) (0.66) (0.54) (-1.56) (-2.65) (-1.57) (-1.87) Cash flow/assets *** *** *** *** (-3.39) (-3.34) (-3.39) (-3.39) (-0.68) (-0.51) (-0.68) (-0.64) NWC/assets *** *** *** *** *** *** *** *** (-30.29) (-30.26) (-30.28) (-30.26) (-28.74) (-28.56) (-28.74) (-28.65) Capex *** *** *** *** *** *** *** *** (-20.36) (-20.14) (-20.31) (-20.19) (-10.15) (-9.67) (-10.15) (-9.92) Leverage *** *** *** *** *** *** *** *** R&D/sales (-31.98) (-31.82) (-31.97) (-31.90) (-36.03) (-35.67) (-36.04) (-35.87) 0.009*** *** 0.009*** 0.009*** 0.058*** *** *** *** (3.85) (3.87) (3.84) (3.87) (4.12) (4.14) (4.11) (4.13) Dividend dummy * * * * * * (-2.15) (-1.93) (-2.16) (-1.91) (-2.48) (-2.34) (-2.48) (-2.32)

58 58 Table 6 (continued) *** *** *** *** *** *** *** *** Acquisition activity (-23.25) (-23.12) (-23.23) (-23.09) (-11.29) (-11.13) (-11.29) (-11.11) Net debt issuance 0.161*** *** 0.161*** 0.160*** 1.101*** 1.093*** 1.103*** 1.099*** Net equity issuance (15.83) (15.83) (15.83) (15.83) (13.05) (13.03) (13.06) (13.05) 0.183*** *** 0.183*** 0.183*** 1.208*** 1.210*** 1.208*** 1.203*** (22.43) (22.41) (22.43) (22.39) (20.57) (20.55) (20.57) (20.50) Loss dummy *** *** *** *** *** *** *** *** (-11.22) (-11.25) (-11.22) (-11.26) (-12.06) (-12.02) (-12.06) (-12.11) Intercept 0.198*** *** 0.197*** 0.198*** *** *** *** *** (19.47) (19.67) (19.38) (19.32) (-19.27) (-18.28) (-19.31) (-19.12) Within R Notes: Cash/Assets defined as cash and marketable securities (data item #1) divided by total assets (data item #6). Log net cash ratio defined as log value of cash and marketable securities (data item #1) divided by (total assets (data item #6)-cash and marketable securities (data item #1)). The sample includes all Compustat firm-year observations from 1980 to 2007 with positive values for the book value of total assets and sales revenue for firms incorporated in the United States. Financial firms (SIC code ) and utilities (SIC codes ) are excluded from the sample, yielding a panel of 118,897 observations for 13,743 unique firms. Missing explanatory values reduce the panel used here to 67,574 firm-year observations for 12,430 unique firms for the firm fixed effects regressions. Taylor prescriptions are calculated based on two types of Taylor rule specifications from Table 2. Policy deviation is the difference between the actual annual average federal funds rates and Taylor prescriptions. Squared policy deviation is the squared value of policy deviation. Cumulative policy deviations are the sum of Taylor rule deviations from the first period up to the current. We calculate the fiscal deficit as the difference between annual federal government current receipts and current expenditures divided by nominal GDP. Credit spread is the difference between the AAA and BBB yields reported by the Federal Reserve. Industry sigma is the average across the two-digit SIC code of the firm cash flow standard deviations for the previous 10 years, and we require at least three observations for the calculation. Market-to-book is the ratio of the market value of assets to the book value of assets i.e. book value of assets (#6) minus the book value of equity (#60) plus the market value of equity (#199* #25) as the numerator of the ratio and the book value of assets (#6) as the denominator. Real size is the logarithm of book assets (#6). Cash flow/assets is calculated as earnings after interest, dividends, and taxes but before depreciation divided by book assets (((#13 #15 #16 #21)/#6). NWC/assets is net working capital (data item #179) minus cash and marketable securities (data item #1) divided by book assets. Capex is the ratio of capital expenditures (data item #128) to the book value of total assets (data item #6). Leverage is the ratio of total debt to the book value of total assets (data item #6), where debt includes long-term debt (data item #9) plus debt in current liabilities (data item #34). R&D/sales is the ratio of research and development expense (data item #46) to sales (data item #12). Dividend dummy is a dummy variable equal to one if the firm paid a common dividend and zero otherwise. Acquisition activity is the ratio of expenditures on acquisitions (data item #129) relative to the book value of total assets (data item #6). Net debt issuance is calculated as annual total debt issuance (data item #111) minus debt retirement (data item #114), divided by the book value of total assets (data item #6). Net equity issuance is calculated as equity sales (data item #108) minus equity purchases (data item #115), divided by the book value of total assets (data item #6). Loss dummy is a dummy variable equal to one if net income (data item #172) is less than zero, and zero otherwise. All variables in dollars are inflation-adjusted to 2007 dollars using the Consumer Price Index. ΔX t is notation for the one-year change, X t- X t-1, where t and (t-1) denote end of fiscal year t and (t-1). 2000s dummy is a dummy variable equal to one if the firm observation is in the fiscal year after 1999, and zero otherwise. In this analysis, we first divide the whole sample into four quartiles each fiscal year based on the real size and define firms in the largest real size quartiles as large firms. Large is a dummy variable equal to one if the firm is in the large real size quartile and zero otherwise. t-statistics based on standard errors robust to clustering by firm and year are reported in parentheses. We report adjusted-r 2 for OLS estimation models and within R 2 for firm fixed effects estimation models. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

59 59 Regressions of Cash Holdings on Taylor Rule Prescription Variables and Controls Model Yearly regression results Quarterly regression results Types of Taylor rule T. 1 T. 2 T. 1 T. 2 T. 1 T. 2 T. 1 T. 2 Dependent variable Cash/Assets Log(Cash/Net Assets) Cash/Assets Log(Cash/Net Assets) Independent variable (5) (6) (7) (8) (5) (6) (7) (8) Taylor prescription large * ** ** * *** *** (-2.06) (-1.49) (-3.21) (-2.94) (-2.11) (-1.70) (-3.92) (-3.97) Taylor prescription ** ** *** * ** *** (-2.67) (-0.50) (2.80) (4.55) (0.48) (2.23) (2.65) (4.94) Funds rate ** * *** *** (3.18) (2.11) (1.46) (0.60) (4.59) (3.72) (0.65) (-0.48) Fiscal deficit *** *** *** *** *** *** *** *** (-4.48) (-4.58) (-7.59) (-8.05) (-5.98) (-6.29) (-7.29) (-7.98) Credit spread *** *** *** *** *** *** *** *** (6.81) (6.69) (11.36) (11.54) (7.49) (7.54) (10.23) (10.26) Industry sigma * *** *** * * (1.95) (2.01) (4.09) (4.05) (-2.05) (-2.06) (-1.88) (-1.92) Market to book *** *** *** *** *** *** *** *** (7.72) (7.78) (10.63) (10.73) (10.15) (10.18) (13.49) (13.54) Real size * * (-0.35) (-0.44) (-1.96) (-2.06) (0.86) (0.74) (-0.80) (-0.97) Cash flow/assets * * *** *** (1.56) (1.55) (2.12) (2.14) (-3.43) (-3.42) (-0.80) (-0.77) NWC/assets *** *** *** *** *** *** *** *** (-34.38) (-34.37) (-30.76) (-30.77) (-30.26) (-30.24) (-28.79) (-28.78) Capex *** *** *** *** *** *** *** *** (-26.99) (-26.96) (-20.68) (-20.68) (-20.24) (-20.20) (-10.15) (-10.13) Leverage *** *** *** *** *** *** *** *** (-33.39) (-33.38) (-36.63) (-36.63) (-31.97) (-31.98) (-36.08) (-36.12) R&D/sales *** *** *** *** *** *** *** *** (3.49) (3.47) (3.74) (3.72) (3.83) (3.82) (4.11) (4.09) Dividend dummy * * * * (-0.12) (-0.20) (-0.83) (-0.88) (-2.09) (-2.14) (-2.36) (-2.41) Acquisition activity *** *** *** *** *** *** *** *** (-25.97) (-25.87) (-16.57) (-16.50) (-23.20) (-23.15) (-11.34) (-11.28) Net debt issuance *** *** *** *** *** *** *** *** (15.37) (15.37) (12.57) (12.58) (15.85) (15.87) (13.07) (13.11)

60 60 Table 7 (continued) *** *** *** *** *** *** *** *** Net equity issuance (23.04) (23.02) (19.01) (19.01) (22.43) (22.43) (20.54) (20.52) Loss dummy *** *** *** *** *** *** *** *** (-17.73) (-17.69) (-14.84) (-14.80) (-11.26) (-11.24) (-12.09) (-12.06) Intercept *** *** *** *** *** *** *** *** (21.13) (21.07) (-18.27) (-18.37) (18.35) (18.42) (-19.56) (-19.77) Adj. R 2 /Within R Note: Cash/Assets defined as cash and marketable securities (data item #1) divided by total assets (data item #6). Log net cash ratio defined as log value of cash and marketable securities (data item #1) divided by (total assets (data item #6)-cash and marketable securities (data item #1)). The sample includes all Compustat firm-year observations from 1980 to 2007 with positive values for the book value of total assets and sales revenue for firms incorporated in the United States. Financial firms (SIC code ) and utilities (SIC codes ) are excluded from the sample, yielding a panel of 118,897 observations for 13,743 unique firms. Missing explanatory values reduce the panel used here to 67,574 firm-year observations for 12,430 unique firms for the firm fixed effects regressions. Taylor prescriptions are calculated based on two types of Taylor rule specifications from Table 2. We calculate the fiscal deficit as the difference between annual federal government current receipts and current expenditures divided by nominal GDP. Credit spread is the difference between the AAA and BBB yields reported by the Federal Reserve. Industry sigma is the average across the two-digit SIC code of the firm cash flow standard deviations for the previous 10 years, and we require at least three observations for the calculation. Market-to-book is the ratio of the market value of assets to the book value of assets i.e. book value of assets (#6) minus the book value of equity (#60) plus the market value of equity (#199* #25) as the numerator of the ratio and the book value of assets (#6) as the denominator. Real size is the logarithm of book assets (#6). Cash flow/assets is calculated as earnings after interest, dividends, and taxes but before depreciation divided by book assets (((#13 #15 #16 #21)/#6). NWC/assets is net working capital (data item #179) minus cash and marketable securities (data item #1) divided by book assets. Capex is the ratio of capital expenditures (data item #128) to the book value of total assets (data item #6). Leverage is the ratio of total debt to the book value of total assets (data item #6), where debt includes long-term debt (data item #9) plus debt in current liabilities (data item #34). R&D/sales is the ratio of research and development expense (data item #46) to sales (data item #12). Dividend dummy is a dummy variable equal to one if the firm paid a common dividend and zero otherwise. Acquisition activity is the ratio of expenditures on acquisitions (data item #129) relative to the book value of total assets (data item #6). Net debt issuance is calculated as annual total debt issuance (data item #111) minus debt retirement (data item #114), divided by the book value of total assets (data item #6). Net equity issuance is calculated as equity sales (data item #108) minus equity purchases (data item #115), divided by the book value of total assets (data item #6). Loss dummy is a dummy variable equal to one if net income (data item #172) is less than zero, and zero otherwise. All variables in dollars are inflation-adjusted to 2007 dollars using the Consumer Price Index. ΔXt is notation for the one-year change, Xt- Xt-1, where t and (t-1) denote end of fiscal year t and (t-1). 2000s dummy is a dummy variable equal to one if the firm observation is in the fiscal year after 1999, and zero otherwise. In this analysis, we first divide the whole sample into four quartiles each fiscal year based on the real size and define firms in the largest real size quartiles as large firms. Large is a dummy variable equal to one if the firm is in the large real size quartile and zero otherwise. t-statistics based on standard errors robust to clustering by firm and year are reported in parentheses. We report adjusted-r 2 for OLS estimation models and within R 2 for firm fixed effects estimation models. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

61 Annual Taylor Rule Prescriptions and the Federal Funds Rate Target Quarterly Taylor Rule Prescriptions and the Federal Funds Rate Target Sources: Bureau of Economic analysis, Congressional budget Office, Bureau of Labor Statistics, Federal Reserve, and my calculations. Taylor Rules calculated as described in text with inflation measured by the 4-quarter rate of change in the CPI and the output gap measured as the log ratio of real GDP to the CBO estimate of potential. We then take average for the quarterly to get my yearly data. 61

62 Annual Temporary Monetary Policy Deviations Quarterly Temporary Monetary Policy Deviations Sources: Bureau of Economic analysis, Congressional budget Office, Bureau of Labor Statistics, Federal Reserve, and my calculations.taylor Rules calculated as described in text with inflation measured by the 4-quarter rate of change in the CPI and the output gap measured as the log ratio of real GDP to the CBO estimate of potential. We then take average for the quarterly to get my yearly data. 62

63 Annual Cumulative Monetary Policy Deviations Quarterly Cumulative Monetary Policy Deviations 63

64 Annual Cumulative Monetary Policy Deviations for Subsample Sources: Bureau of Economic analysis, Congressional budget Office, Bureau of Labor Statistics, Federal Reserve, and my calculations.taylor Rules calculated as described in text with inflation measured by the 4-quarter rate of change in the CPI and the output gap measured as the log ratio of real GDP to the CBO estimate of potential. We then take average for the quarterly to get my yearly data. 64