# Financial Frictions and the Stock Price Reaction to Monetary Policy

Financial Frictions and the Stock Price Reaction to Monetary Policy
Ozdagli, Ali K.
2017-09-18 00:00:00
Abstract I show that the stock prices of firms subject to greater information frictions have a weaker reaction to monetary policy. The claim is robust to a broad set of proxies for financial constraints and information frictions. Moreover, I use the Enron accounting scandal and Arthur Andersen’s demise as a large exogenous shock, temporarily raising other Andersen clients’ information frictions and, thereby, their financial constraints. The scandal’s disclosure lowered Andersen’s clients’ stock price sensitivity to monetary policy to about half that of other firms. Received February 12, 2016; editorial decision August 22, 2017 by Editor Robin Greenwood.Authors have furnished an Internet Appendix, which is available on the Oxford University PressWeb site next to the link to the final published paper online. The impact of monetary policy on firms’ market values has been recognized at least since Tobin (1969) and Modigliani (1971). However, we still know very little about the channels through which stock prices respond to monetary policy. This paper finds a new link between stock prices and monetary policy: stock prices of firms subject to greater information frictions are less responsive to monetary policy shocks than other firms. This empirical pattern conforms with the fact that firms with greater information frictions have lower leverage and leverage amplifies the stock price reaction to monetary policy. In particular, these firms’ lower leverage implies that the fluctuations in the cost of debt caused by monetary policy may affect their stock prices less. I start by showing that, among the many channels that connect monetary policy and stock prices, financial constraints play a significant role. For this purpose, I use policy surprises based on federal funds futures prices on Federal Open Market Committee (FOMC) announcement dates. I find that, on the dates of the scheduled FOMC announcements, firms that are more constrained according to broad financial constraint measures, such as the Whited and Wu (2006) index and the Hadlock and Pierce (2010) index, have stock prices that are about one-third less sensitive to monetary policy surprises. Furthermore, the median leverage of these firms is less than half the median leverage of unconstrained firms. Although financial constraints can result from many sources, they are often a manifestation of information frictions. Indeed, Tirole (2006) classifies a vast portion of capital structure theories under adverse selection or moral hazard, both of which are symptoms of information frictions. Moreover, previous studies often justify their use of a particular financial constraint proxy by citing information frictions. For example, Gertler and Gilchrist (1994) argue that smaller firms are more constrained because “the informational frictions that add to the costs of external finance apply mainly to [$$\ldots$$] smaller firms.” Therefore, I complement my results by looking at commonly used measures of information frictions, including whether the firm is unrated, has high R&D expenses, or has positive accruals and hence is more likely to engage in earnings management. The results are largely consistent with the results obtained by using the Hadlock and Pierce (HP) and Whited and Wu (WW) indices. For example, while the average firm’s stock price increases by about 3% in response to a 1-percentage-point expansionary policy rate surprise, unrated firms respond by about 1.30 percentage points less than others. Similar to the case of the HP and WW indices, the median leverage of unrated firms is lower than that of rated firms, about 0.08 versus 0.31. This pattern conforms with the intuition that constrained firms’ lower leverage leads to lower sensitivity of stock prices to the fluctuations in the cost of debt caused by monetary policy. Although these results are suggestive, they do not cleanly isolate the channel through which financial frictions drive the stock price sensitivity to monetary policy. The plethora of financial constraint indices use different firm characteristics, which blurs the precise mechanism through which the financial constraints operate. For example, the proxies for financial constraints may actually reflect faster growth opportunities rather than firms’ ability to finance their projects (Farre-Mensa and Ljungqvist (2015)). Then the estimates are biased toward zero because growth options have higher cash flow duration and implicit leverage (Berk, Green, and Naik (2004)), which can make the stock price of a firm with bigger growth opportunities more responsive to monetary policy. To improve the identification of the relation between financial frictions and policy sensitivity and to isolate the channel through which these frictions operate, I use the Enron accounting scandal of 2001 and the resultant demise of its auditing firm, Arthur Andersen, as an exogenous large shock to the information frictions of other Arthur Andersen clients relative to the clients of other auditors.1 Because information frictions are greatly reduced by the independent auditing of firms’ financial statements, a sudden loss in the credibility of a major auditing firm is an important and long-lasting shock to its clients. Even lenders with specialized information, such as banks, benefit from independent auditing because bank loans have covenants based on financial statement variables. Accordingly, Graham, Li, and Qiu (2008) find that financial restatements lead to increased bank loan rates and tighter covenants for the restating firms. This identification approach is further supported by the fact that Arthur Andersen clients have experienced a large and statistically significant decrease in their leverage relative to the clients of other auditors in the years following the scandal. Although a firm’s capital structure adjusts slowly, stock prices are forward looking and incorporate any new information immediately. Therefore, I look at how stock prices respond to monetary policy announcements in 2001. As a first approach, the paper focuses on two scheduled FOMC announcement dates with similar policy actions, May 15 and November 6. The first announcement was made before the widespread accusations regarding Enron’s accounting practices and the second one was made right after these accusations, following the SEC inquiry into Enron’s financial statements on October 22, 2001. On both dates, the FOMC reduced rates by 50 basis points (bp), and federal funds futures indicate that about 10 bp of each reduction came to the market as a big surprise. As a result, the S&P 500 index has increased swiftly by 80 basis points in the hour following the announcements, with considerable variation across stocks. The large surprise and the considerable heterogeneity in stocks’ response help us significantly in identifying the effect of the policy on stock prices. In accordance with my previous results from information friction proxies, I find that the stock prices of Arthur Andersen clients responded by about 50 to 70 bp less than the stock prices of other firms on November 6. This result is robust across many specifications, including various controls, as well as an instrumental variable approach and a placebo test. Moreover, the effect is particularly strong among firms with no ratings, likely reflecting that a reliable financial statement is a more important information source for an opaque firm. A second complementary panel data analysis uses Enron’s expected default frequency (Moody’s EDF) as a proxy for investors’ beliefs about the scandal. This analysis uses monetary policy surprises on the eight scheduled announcement dates in 2001, allowing us to control directly for potentially confounding effects of other macroeconomic events and to address potential concerns due to minor differences between the monetary policy surprises of May 15 and November 6. The policy surprises in 2001 are large, with an average absolute surprise of 4.75 bp compared to 3.35 bp for the rest of the sample used in previous regressions, which makes this period particularly useful for identification as well. Consistent with previous results, the stock prices of Arthur Andersen clients responded by about 50 bp less than the stock prices of other firms to a 10-bp surprise reduction in the federal funds target rate toward the end of the scandal. This effect is large, considering that, on average, a 10-bp surprise decrease in the federal funds target rate led to a 100-bp increase in daily stock prices in 2001. Similarly, this result is much stronger among firms with no ratings, high R&D spending, and positive accruals, further highlighting the importance of reliable financial statements for opaque firms. Finally, the results regarding the importance of the reliability of financial statements are generalized by using financial restatements of different firms from 1997 to 2006, with the average post-restatement effect being about 40 bp for a 10-bp policy surprise. While the literature agrees that monetary policy has a significant effect on stock prices, we have known very little about the transmission channels.2 This paper cleanly identifies the channel through which information frictions operate and thereby contributes to the literature that studies the nexus of financial frictions, monetary policy, and stock prices. There are relatively few papers in this nexus and they reach mixed conclusions because of various identification problems that are resolved in my paper. Perez-Quiros and Timmermann (2000) use the lagged change in the monetary base as a proxy for monetary policy decisions and find that small firms’ stock prices react more strongly to monetary policy. Lamont, Polk, and Saá-Requejo (2001) recognize that modern monetary policy works through interest rates rather than through the monetary base and hence study the change in the federal funds rate and the discount window rate. They find no evidence that the relative performance of constrained firms reflects monetary policy, credit conditions, or business cycles. This result is probably not surprising because the change in the interest rate has both an anticipated and an unanticipated component and stock prices are unlikely to respond to anticipated changes in monetary policy. Therefore, I study the reaction of stock prices to the unanticipated component of the policy rate changes. While the previous literature singles out firm size as a financial constraint proxy, I use multiple different types of proxies to derive my conclusions. More importantly, unlike the previous literature, I provide a clean identification using a natural experiment with an exogenous variation in an important financial friction, the unreliability of financial statements. The improved identification in this paper allows a clear demonstration of how financial frictions may affect the policy sensitivity of stock prices through changes in the firms’ capital structure, a finding that is also supported by a simple theoretical analysis based on the well-known framework of Bernanke, Gertler, and Gilchrist (1999). As a result, the paper extends our knowledge of monetary policy transmission to the real economy. The next section starts with an overview of the determinants of the stock price sensitivity to monetary policy. In the subsequent sections, I focus on isolating the effect of financial and information frictions. I conclude with a discussion of potential channels that may explain why firms subject to greater information frictions have stock prices that are less sensitive to monetary policy and show that the lower leverage of these firms is the mechanism that stands out. 1. The Role of Financial Frictions and Other Channels in Monetary Policy Transmission Monetary policy has a large effect on stock prices and this effect varies significantly across firms. The literature has come up with multiple channels that can explain this cross-sectional variation. This section discusses these channels, highlighting the importance of information frictions as one of the main channels of monetary policy transmission. 1.1 Measuring monetary policy surprises and policy sensitivities Estimating the response of stock prices to monetary policy actions is complicated by the fact that the market is unlikely to respond to policy actions that are already anticipated because this information is already incorporated in prices prior to the FOMC announcements. To address this complexity, we need to calculate the unexpected (surprise) component of monetary policy changes. Monetary policy surprises in this paper are measured by following the method in Kuttner (2001) and Bernanke and Kuttner (2005) (henceforth BK 2005), who obtain the unexpected changes in the federal funds target rate on FOMC announcement dates using the expectations embedded in federal funds futures contracts. This method first calculates the change in the rate implied by the corresponding futures contract, given by 100 minus the futures contract price. This result is then scaled by a factor associated with the number of days of the month in which the event occurred, because the payoff of the contract is determined by the average realized federal funds effective rate during the month. Accordingly, the unexpected target-rate change for an event taking place on day $$d$$ of month $$m$$ is given by \begin{equation} \Delta i^{u}=\frac{D}{D-d}(f_{m,d}^{0}-f_{m,d-1}^{0}), \end{equation} (1) where $$f_{m,d}^{0}-f_{m,d-1}^{0}$$ is the change in the current-month implied futures rate, and $$D$$ is the number of days in the month. To suppress the end-of-month noise in the federal funds rate, the unscaled change in the implied futures rate is used as the measure of the target-rate surprise when the event occurs during the last three days of a month. If the event happens on the first day of a month, $$f_{m-1,D}^{1}$$ is used instead of $$f_{m,d-1}^{0}$$. The resultant data for these policy surprises are publicly available from Kenneth Kuttner’s Web page (http://econ.williams.edu/profile/knk1/). I use only the FOMC announcements made on dates scheduled in advance because unscheduled meetings tend to be reactive or endogenous, such as the FOMC announcement following the Long-Term Capital Management (LTCM) crisis. This choice minimizes the contamination of monetary policy surprises by other macroeconomic news. This approach also minimizes delays in the incorporation of information to stock prices, thereby improving identification, because the dates of scheduled meetings are known well in advance. The monetary policy surprise throughout this paper is the unexpected change in the federal funds target rate multiplied by $$-1$$, so that a positive surprise is expansionary, \begin{equation} PolicySurprise=-\Delta i^{u}=-\frac{D}{D-d}(f_{m,d}^{0}-f_{m,d-1}^{0})\text{.} \end{equation} (2) I focus on the period from 1994 to mid-2008 because the federal funds target rate was the explicit policy instrument during this time period until the zero lower bound. Given the Fed’s recent move toward policy normalization, the federal funds target rate will be the relevant policy measure for the future as well. This approach is preferable because federal funds futures outperform target-rate forecasts based on other financial market instruments or on alternative methods, such as sophisticated time-series specifications and monetary policy rules (Evans (1998); Gürkaynak, Sack, and Swanson (2007)). Another advantage of looking at one-day changes in near-dated federal funds futures is that federal funds futures do not exhibit predictable time-varying risk premiums (and forecast errors) over daily frequencies, as discussed in Piazzesi and Swanson (2008). BK (2005) present a simple method to calculate policy sensitivities using $$PolicySurprise$$. In particular, we can measure the policy sensitivity, $$\beta $$, from the following regression: \begin{equation} r_{t}=\alpha +\beta \times \textit{PolicySurprise}_{t}+\varepsilon _{t}, \label{eq: BK_regression} \end{equation} (3) where date $$t$$ is the date of a scheduled FOMC meeting. The return and policy surprise on this date are $$r_{t}$$ and $$PolicySurprise_{t}$$, both measured in percentage points. Figure 1 shows the cross-sectional distribution of the monetary policy sensitivity. The reaction of 90% of the firms to a $$1$$-percentage-point policy surprise lies between $$-40.2$$ and $$44.6$$ percentage points, suggesting large differences across firms. Figure 1 View largeDownload slide Distribution of monetary policy sensitivity The figure gives the cross-sectional distribution of how stock prices react to monetary policy (5%$$-$$95% range). Sensitivity comes from equation (3). The time period goes from beginning of 1994 to mid-2008 because federal funds target rate was the explicit monetary policy instrument during this time period. Starting in February 1994, the FOMC’s policy of announcing new target rates at prescheduled dates virtually eliminated the timing ambiguity associated with rate changes prior to this date. After mid-2008, the Federal Reserve switched from announcing a specific target rate to announcing a range for the target rate. The FOMC announcement on March 18, 2008, is dropped as an outlier because on that day, the S&P 500 index increased by about 4% despite a 17-bp contractionary policy surprise, reflecting the positive news about JP Morgan’s purchase of Bearn Stearns. Including this date changes policy sensitivity of the CRSP value-weighted index from 3.2 percentage points, in line with earlier estimates of Bernanke and Kuttner (2005), to 1.3 percentage points. Figure 1 View largeDownload slide Distribution of monetary policy sensitivity The figure gives the cross-sectional distribution of how stock prices react to monetary policy (5%$$-$$95% range). Sensitivity comes from equation (3). The time period goes from beginning of 1994 to mid-2008 because federal funds target rate was the explicit monetary policy instrument during this time period. Starting in February 1994, the FOMC’s policy of announcing new target rates at prescheduled dates virtually eliminated the timing ambiguity associated with rate changes prior to this date. After mid-2008, the Federal Reserve switched from announcing a specific target rate to announcing a range for the target rate. The FOMC announcement on March 18, 2008, is dropped as an outlier because on that day, the S&P 500 index increased by about 4% despite a 17-bp contractionary policy surprise, reflecting the positive news about JP Morgan’s purchase of Bearn Stearns. Including this date changes policy sensitivity of the CRSP value-weighted index from 3.2 percentage points, in line with earlier estimates of Bernanke and Kuttner (2005), to 1.3 percentage points. 1.2 Channels of monetary policy transmission to stock prices In the rest of the paper, we try to understand the cross-sectional differences in policy sensitivity by studying the effects of different firm characteristics summarized in Table 1. They key point is that while industry and cash flow characteristics play a significant role, financial frictions are at the heart of the transmission of monetary policy to stock prices. Table 1 Description of firm-level variables Variable Construction (Compustat item code) Accruals Sloan (1996) accruals, using current assets $$(\text{ACT})$$, cash and short-term investments (CHE), current liabilities (LCT), debt in current liabilities (DLC), income taxes payable (TXP), depreciation and amortization (DP), and total assets (AT), using the following formula where $$\Delta$$ refers to change, $$((\Delta\text{ACT} - \Delta\text{CHE})-(\Delta\text{LCT} -\Delta\text{DLC}-\Delta\text{TXP})$$$$ -\text{DP})/((\text{AT}+\text{lagged AT})/2)$$ Book-to-market $$(\text{CEQ}+\text{TXDITC})$$ divided by the product of the stock price and shares outstanding from CRSP as of December of the same year CAPM-implied sensitivity CAPM beta multiplied by the policy sensitivity of the CRSP value-weighted returns (3.2) Cash flow coefficient of variation The standard deviation of quarterly cash flows (EBITDAQ) over the last five years normalized by the absolute value of the average cash flows during the same window Cash flow duration Duration of cash flows from assets as measured in Weber (Forthcoming) Cash flow estimated sensitivity The regression coefficient that comes from regressing change in quarterly cash flows (EBITDAQ), normalized by lagged assets (ATQ), on the change in 3-month LIBOR over the same quarter Financial constraint index (KZ) $$-1.001909[(\text{IB}+\text{DP})/\text{lagged PPENT}] + 0.2826389[ (\text{AT} + \text{PRCC_F}\times\text{CSHO} - \text{CEQ} - \text{TXDB})/\text{AT}] + 3.139193[(\text{DLTT} + \text{DLC})/(\text{DLTT} + \text{DLC} + \text{SEQ})] - 39.3678[(\text{DVC} + \text{DVP})/\text{lagged PPENT}] - 1.314759[\text{CHE}/\text{lagged PPENT}]$$. See Farre-Mensa and Ljungqvist (2015) for details Financial constraint index (HP) $$-0.737\text{Size} + 0.043\text{Size}^2 - 0.040\text{Age}$$, where Size equals the log of inflation-adjusted Compustat item AT (in 2004 dollars), and Age is the number of years the firm is listed with a nonmissing stock Price or assets on Compustat. Size is capped at (the log of) $\$$4.5 billion and Age is capped at 37 years. See Farre-Mensa and Ljungqvist (2015) for details Financial constraint index (WW) $$-0.091 [(\text{IB} + \text{DP})/\text{AT}] - 0.062[$$indicator set to one if $$\text{DVC} + \text{DVP}$$ is positive, and zero otherwise]$$ + 0.021[\text{DLTT}/\text{AT}] - 0.044[\log(\text{AT})]$$ + $$ 0.102[$$average SIC 3-digit industry sales growth each year]$$ - 0.035[\text{sales growth}]$$. See Farre-Mensa and Ljungqvist (2015) for details. Sales growth is Logarithm of total sales (SALEQ) during quarters ‘$$t$$’ to ‘$$t+s$$’ $$-$$ logarithm of total sales during quarters “$$t-s-1$$” to “$$t-1$$” (sales are deflated to base year 2000) Market value of equity The product of the stock price and shares outstanding as of December, deflated to base year 2000 R&D spending The ratio of R&D spending (XRD) divided by total assets (AT) Unrated A dummy variable that takes the value of one if the firm does not have a long-term credit rating from S&P at any month in a given year, and zero otherwise. See, for example, Avramov et al. (2007) and Colla, Ippolito, and Li (2013). Variable Construction (Compustat item code) Accruals Sloan (1996) accruals, using current assets $$(\text{ACT})$$, cash and short-term investments (CHE), current liabilities (LCT), debt in current liabilities (DLC), income taxes payable (TXP), depreciation and amortization (DP), and total assets (AT), using the following formula where $$\Delta$$ refers to change, $$((\Delta\text{ACT} - \Delta\text{CHE})-(\Delta\text{LCT} -\Delta\text{DLC}-\Delta\text{TXP})$$$$ -\text{DP})/((\text{AT}+\text{lagged AT})/2)$$ Book-to-market $$(\text{CEQ}+\text{TXDITC})$$ divided by the product of the stock price and shares outstanding from CRSP as of December of the same year CAPM-implied sensitivity CAPM beta multiplied by the policy sensitivity of the CRSP value-weighted returns (3.2) Cash flow coefficient of variation The standard deviation of quarterly cash flows (EBITDAQ) over the last five years normalized by the absolute value of the average cash flows during the same window Cash flow duration Duration of cash flows from assets as measured in Weber (Forthcoming) Cash flow estimated sensitivity The regression coefficient that comes from regressing change in quarterly cash flows (EBITDAQ), normalized by lagged assets (ATQ), on the change in 3-month LIBOR over the same quarter Financial constraint index (KZ) $$-1.001909[(\text{IB}+\text{DP})/\text{lagged PPENT}] + 0.2826389[ (\text{AT} + \text{PRCC_F}\times\text{CSHO} - \text{CEQ} - \text{TXDB})/\text{AT}] + 3.139193[(\text{DLTT} + \text{DLC})/(\text{DLTT} + \text{DLC} + \text{SEQ})] - 39.3678[(\text{DVC} + \text{DVP})/\text{lagged PPENT}] - 1.314759[\text{CHE}/\text{lagged PPENT}]$$. See Farre-Mensa and Ljungqvist (2015) for details Financial constraint index (HP) $$-0.737\text{Size} + 0.043\text{Size}^2 - 0.040\text{Age}$$, where Size equals the log of inflation-adjusted Compustat item AT (in 2004 dollars), and Age is the number of years the firm is listed with a nonmissing stock Price or assets on Compustat. Size is capped at (the log of) $\$$4.5 billion and Age is capped at 37 years. See Farre-Mensa and Ljungqvist (2015) for details Financial constraint index (WW) $$-0.091 [(\text{IB} + \text{DP})/\text{AT}] - 0.062[$$indicator set to one if $$\text{DVC} + \text{DVP}$$ is positive, and zero otherwise]$$ + 0.021[\text{DLTT}/\text{AT}] - 0.044[\log(\text{AT})]$$ + $$ 0.102[$$average SIC 3-digit industry sales growth each year]$$ - 0.035[\text{sales growth}]$$. See Farre-Mensa and Ljungqvist (2015) for details. Sales growth is Logarithm of total sales (SALEQ) during quarters ‘$$t$$’ to ‘$$t+s$$’ $$-$$ logarithm of total sales during quarters “$$t-s-1$$” to “$$t-1$$” (sales are deflated to base year 2000) Market value of equity The product of the stock price and shares outstanding as of December, deflated to base year 2000 R&D spending The ratio of R&D spending (XRD) divided by total assets (AT) Unrated A dummy variable that takes the value of one if the firm does not have a long-term credit rating from S&P at any month in a given year, and zero otherwise. See, for example, Avramov et al. (2007) and Colla, Ippolito, and Li (2013). Summary statistics Variable Number of observations Number of Mean SD Variable Number of observations Number of Mean SD CF variation 341,117 53.18 114.30 Accruals 316,760 –0.03 0.10 CF duration 288,797 19.54 5.14 KZ index 242,525 –10.01 178.46 CF sensitivity 341,144 0.10 7.03 HP index 446,696 –3.32 0.69 Book-to-market 362,841 0.61 0.73 WW index 256,528 –0.27 0.77 log(Market val) 445,923 1.39 1.78 R&D spending 211,592 0.08 0.15 CAPM-impl. sensitivity 446,653 2.47 3.01 Unrated 447,173 0.71 0.45 Summary statistics Variable Number of observations Number of Mean SD Variable Number of observations Number of Mean SD CF variation 341,117 53.18 114.30 Accruals 316,760 –0.03 0.10 CF duration 288,797 19.54 5.14 KZ index 242,525 –10.01 178.46 CF sensitivity 341,144 0.10 7.03 HP index 446,696 –3.32 0.69 Book-to-market 362,841 0.61 0.73 WW index 256,528 –0.27 0.77 log(Market val) 445,923 1.39 1.78 R&D spending 211,592 0.08 0.15 CAPM-impl. sensitivity 446,653 2.47 3.01 Unrated 447,173 0.71 0.45 Summary statistics are given for those observations for which the value of the variable is nonmissing and the firm has a nonmissing industry code. For example, the HP index uses only firm size and age, which are available for most firms, and therefore it has more observations than the WW index, which has six terms that use seven Compustat variables, some of which can be missing for a given firm. Following previous literature, the regressions assign firms to discrete groups based on the value of the financial constraint measure every year. This approach is robust to outliers and linear transformations. For example, like in Farre-Mensa and Ljungqvist (2015), the Whited-Wu index does not include a constant and the deflator for $$\log(\text{AT})$$, which are irrelevant for the ranking. Table 1 Description of firm-level variables Variable Construction (Compustat item code) Accruals Sloan (1996) accruals, using current assets $$(\text{ACT})$$, cash and short-term investments (CHE), current liabilities (LCT), debt in current liabilities (DLC), income taxes payable (TXP), depreciation and amortization (DP), and total assets (AT), using the following formula where $$\Delta$$ refers to change, $$((\Delta\text{ACT} - \Delta\text{CHE})-(\Delta\text{LCT} -\Delta\text{DLC}-\Delta\text{TXP})$$$$ -\text{DP})/((\text{AT}+\text{lagged AT})/2)$$ Book-to-market $$(\text{CEQ}+\text{TXDITC})$$ divided by the product of the stock price and shares outstanding from CRSP as of December of the same year CAPM-implied sensitivity CAPM beta multiplied by the policy sensitivity of the CRSP value-weighted returns (3.2) Cash flow coefficient of variation The standard deviation of quarterly cash flows (EBITDAQ) over the last five years normalized by the absolute value of the average cash flows during the same window Cash flow duration Duration of cash flows from assets as measured in Weber (Forthcoming) Cash flow estimated sensitivity The regression coefficient that comes from regressing change in quarterly cash flows (EBITDAQ), normalized by lagged assets (ATQ), on the change in 3-month LIBOR over the same quarter Financial constraint index (KZ) $$-1.001909[(\text{IB}+\text{DP})/\text{lagged PPENT}] + 0.2826389[ (\text{AT} + \text{PRCC_F}\times\text{CSHO} - \text{CEQ} - \text{TXDB})/\text{AT}] + 3.139193[(\text{DLTT} + \text{DLC})/(\text{DLTT} + \text{DLC} + \text{SEQ})] - 39.3678[(\text{DVC} + \text{DVP})/\text{lagged PPENT}] - 1.314759[\text{CHE}/\text{lagged PPENT}]$$. See Farre-Mensa and Ljungqvist (2015) for details Financial constraint index (HP) $$-0.737\text{Size} + 0.043\text{Size}^2 - 0.040\text{Age}$$, where Size equals the log of inflation-adjusted Compustat item AT (in 2004 dollars), and Age is the number of years the firm is listed with a nonmissing stock Price or assets on Compustat. Size is capped at (the log of) $\$$4.5 billion and Age is capped at 37 years. See Farre-Mensa and Ljungqvist (2015) for details Financial constraint index (WW) $$-0.091 [(\text{IB} + \text{DP})/\text{AT}] - 0.062[$$indicator set to one if $$\text{DVC} + \text{DVP}$$ is positive, and zero otherwise]$$ + 0.021[\text{DLTT}/\text{AT}] - 0.044[\log(\text{AT})]$$ + $$ 0.102[$$average SIC 3-digit industry sales growth each year]$$ - 0.035[\text{sales growth}]$$. See Farre-Mensa and Ljungqvist (2015) for details. Sales growth is Logarithm of total sales (SALEQ) during quarters ‘$$t$$’ to ‘$$t+s$$’ $$-$$ logarithm of total sales during quarters “$$t-s-1$$” to “$$t-1$$” (sales are deflated to base year 2000) Market value of equity The product of the stock price and shares outstanding as of December, deflated to base year 2000 R&D spending The ratio of R&D spending (XRD) divided by total assets (AT) Unrated A dummy variable that takes the value of one if the firm does not have a long-term credit rating from S&P at any month in a given year, and zero otherwise. See, for example, Avramov et al. (2007) and Colla, Ippolito, and Li (2013). Variable Construction (Compustat item code) Accruals Sloan (1996) accruals, using current assets $$(\text{ACT})$$, cash and short-term investments (CHE), current liabilities (LCT), debt in current liabilities (DLC), income taxes payable (TXP), depreciation and amortization (DP), and total assets (AT), using the following formula where $$\Delta$$ refers to change, $$((\Delta\text{ACT} - \Delta\text{CHE})-(\Delta\text{LCT} -\Delta\text{DLC}-\Delta\text{TXP})$$$$ -\text{DP})/((\text{AT}+\text{lagged AT})/2)$$ Book-to-market $$(\text{CEQ}+\text{TXDITC})$$ divided by the product of the stock price and shares outstanding from CRSP as of December of the same year CAPM-implied sensitivity CAPM beta multiplied by the policy sensitivity of the CRSP value-weighted returns (3.2) Cash flow coefficient of variation The standard deviation of quarterly cash flows (EBITDAQ) over the last five years normalized by the absolute value of the average cash flows during the same window Cash flow duration Duration of cash flows from assets as measured in Weber (Forthcoming) Cash flow estimated sensitivity The regression coefficient that comes from regressing change in quarterly cash flows (EBITDAQ), normalized by lagged assets (ATQ), on the change in 3-month LIBOR over the same quarter Financial constraint index (KZ) $$-1.001909[(\text{IB}+\text{DP})/\text{lagged PPENT}] + 0.2826389[ (\text{AT} + \text{PRCC_F}\times\text{CSHO} - \text{CEQ} - \text{TXDB})/\text{AT}] + 3.139193[(\text{DLTT} + \text{DLC})/(\text{DLTT} + \text{DLC} + \text{SEQ})] - 39.3678[(\text{DVC} + \text{DVP})/\text{lagged PPENT}] - 1.314759[\text{CHE}/\text{lagged PPENT}]$$. See Farre-Mensa and Ljungqvist (2015) for details Financial constraint index (HP) $$-0.737\text{Size} + 0.043\text{Size}^2 - 0.040\text{Age}$$, where Size equals the log of inflation-adjusted Compustat item AT (in 2004 dollars), and Age is the number of years the firm is listed with a nonmissing stock Price or assets on Compustat. Size is capped at (the log of) $\$$4.5 billion and Age is capped at 37 years. See Farre-Mensa and Ljungqvist (2015) for details Financial constraint index (WW) $$-0.091 [(\text{IB} + \text{DP})/\text{AT}] - 0.062[$$indicator set to one if $$\text{DVC} + \text{DVP}$$ is positive, and zero otherwise]$$ + 0.021[\text{DLTT}/\text{AT}] - 0.044[\log(\text{AT})]$$ + $$ 0.102[$$average SIC 3-digit industry sales growth each year]$$ - 0.035[\text{sales growth}]$$. See Farre-Mensa and Ljungqvist (2015) for details. Sales growth is Logarithm of total sales (SALEQ) during quarters ‘$$t$$’ to ‘$$t+s$$’ $$-$$ logarithm of total sales during quarters “$$t-s-1$$” to “$$t-1$$” (sales are deflated to base year 2000) Market value of equity The product of the stock price and shares outstanding as of December, deflated to base year 2000 R&D spending The ratio of R&D spending (XRD) divided by total assets (AT) Unrated A dummy variable that takes the value of one if the firm does not have a long-term credit rating from S&P at any month in a given year, and zero otherwise. See, for example, Avramov et al. (2007) and Colla, Ippolito, and Li (2013). Summary statistics Variable Number of observations Number of Mean SD Variable Number of observations Number of Mean SD CF variation 341,117 53.18 114.30 Accruals 316,760 –0.03 0.10 CF duration 288,797 19.54 5.14 KZ index 242,525 –10.01 178.46 CF sensitivity 341,144 0.10 7.03 HP index 446,696 –3.32 0.69 Book-to-market 362,841 0.61 0.73 WW index 256,528 –0.27 0.77 log(Market val) 445,923 1.39 1.78 R&D spending 211,592 0.08 0.15 CAPM-impl. sensitivity 446,653 2.47 3.01 Unrated 447,173 0.71 0.45 Summary statistics Variable Number of observations Number of Mean SD Variable Number of observations Number of Mean SD CF variation 341,117 53.18 114.30 Accruals 316,760 –0.03 0.10 CF duration 288,797 19.54 5.14 KZ index 242,525 –10.01 178.46 CF sensitivity 341,144 0.10 7.03 HP index 446,696 –3.32 0.69 Book-to-market 362,841 0.61 0.73 WW index 256,528 –0.27 0.77 log(Market val) 445,923 1.39 1.78 R&D spending 211,592 0.08 0.15 CAPM-impl. sensitivity 446,653 2.47 3.01 Unrated 447,173 0.71 0.45 Summary statistics are given for those observations for which the value of the variable is nonmissing and the firm has a nonmissing industry code. For example, the HP index uses only firm size and age, which are available for most firms, and therefore it has more observations than the WW index, which has six terms that use seven Compustat variables, some of which can be missing for a given firm. Following previous literature, the regressions assign firms to discrete groups based on the value of the financial constraint measure every year. This approach is robust to outliers and linear transformations. For example, like in Farre-Mensa and Ljungqvist (2015), the Whited-Wu index does not include a constant and the deflator for $$\log(\text{AT})$$, which are irrelevant for the ranking. BK (2005) show that stock prices of firms in more procyclical industries are more sensitive than the stock prices of other firms to policy surprises, consistent with the idea that some of the cross-sectional variation is driven by demand and industry characteristics. They also find that the estimated policy sensitivities of industry portfolios are very close to these portfolios’ policy sensitivities implied by the capital asset pricing model (CAPM).3 This pattern naturally raises the question of whether the power of CAPM-implied policy sensitivities goes beyond the industry portfolios. To address this issue, I use firm-level data in the following regression: \begin{eqnarray} r_{it} &=&\alpha +\beta \times \textit{PolicySurprise}_{t}+\gamma \times \textit{CAPMimpliedSensitivity}_{i} \nonumber \label{eq: CAPMimplied} \\ &&+\delta \times \textit{CAPMimpliedSensitivity}_{i}\times PolicySurprise_{t} \\ &&+\textit{Controls}+e_{it}, \nonumber \end{eqnarray} (4) where date $$t$$ is the date of a scheduled FOMC meeting, $$r_{it}$$ is the return of stock $$i$$ on date $$t$$, and $$PolicySurprise_{t}$$ is the monetary policy surprise on that date. Similar to BK (2005), $$CAPMimpliedSensitivity_{i}$$ is the CAPM beta of stock $$i$$ on FOMC dates multiplied by the estimated policy sensitivity of the market return ($$3.2$$) which comes from running regression (3) with the CRSP value-weighted return. Regression (4) also includes the 10 Fama-French industry fixed effects and their interaction with $$PolicySurprise_{t}$$ as control variables ($$Controls$$). If CAPM-implied sensitivities can still capture the cross-sectional variation in sensitivities at the firm-level after controlling for industry effects, we should see that $$\delta $$ should be statistically significant and close to $$1 $$. However, as we see in the first column of Table 2, the estimate of $$\delta $$ is very small and statistically insignificant. Since the predictive power of CAPM for the cross-sectional differences in policy sensitivity does not reach beyond industry portfolios, I directly control for industry effects in the remainder of this section.4 Table 2 The relationship of CAPM-implied sensitivity and cash flow characteristics with the stock price sensitivity to monetary policy All firms, scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{it-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ Variables (1) CAPM implied sensitivity (2) Cash flow estimated sensitivity (3) Cash flow duration (4) Cash flow variation (5) Book-to-market (6) log market value PolicySurprise 2.04*** 2.08*** 2.36*** 2.08*** 2.36*** 2.31*** (0.49) (0.33) (0.36) (0.33) (0.35) (0.34) PolicySurprise*Characteristic 0.13 0.01 0.48*** 0.29*** –0.14 1.07*** (0.18) (0.07) (0.17) (0.09) (0.14) (0.10) Characteristic 0.02* –0.00 –0.06*** 0.00 0.09*** –0.21*** (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Constant 0.12*** 0.48*** –3.66*** 0.48*** –1.03*** –1.13*** (0.03) (0.18) (0.02) (0.18) (0.23) (0.25) Number of observations 446,653 341,144 288,797 341,117 362,841 445,923 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 Number of stocks 10,986 8,044 7,246 8,043 9,587 11,320 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes All firms, scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{it-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ Variables (1) CAPM implied sensitivity (2) Cash flow estimated sensitivity (3) Cash flow duration (4) Cash flow variation (5) Book-to-market (6) log market value PolicySurprise 2.04*** 2.08*** 2.36*** 2.08*** 2.36*** 2.31*** (0.49) (0.33) (0.36) (0.33) (0.35) (0.34) PolicySurprise*Characteristic 0.13 0.01 0.48*** 0.29*** –0.14 1.07*** (0.18) (0.07) (0.17) (0.09) (0.14) (0.10) Characteristic 0.02* –0.00 –0.06*** 0.00 0.09*** –0.21*** (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Constant 0.12*** 0.48*** –3.66*** 0.48*** –1.03*** –1.13*** (0.03) (0.18) (0.02) (0.18) (0.23) (0.25) Number of observations 446,653 341,144 288,797 341,117 362,841 445,923 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 Number of stocks 10,986 8,044 7,246 8,043 9,587 11,320 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Regressions 2–6 include firm fixed effects. Errors in parentheses are clustered at the firm level. $$^{***}$$ p<.01, $$^{**}$$ p<.05, $$^*$$ p<.1. The dependent variable is the stock returns of publicly listed companies on FOMC announcement dates. PolicySurprise is the unexpected federal funds target rate change on a scheduled FOMC meeting; it is scaled so that positive surprises indicate expansionary shocks. Both returns and PolicySurprise are expressed in percentage points. The coefficient of PolicySurprise refers to the base industry which is nondurables. CAPM-implied sensitivity is the product of the CAPM beta of the stock and the policy sensitivity of CRSP value-weighted returns. The characteristics are defined in Table 1; they are standardized to zero mean and unit variance. The sample includes firms traded in AMEX, NYSE, and NASDAQ. To ensure the liquidity of the stocks, penny stocks are dropped, following Amihud (2002), who defines penny stocks as stocks with a price less than $\$$5. Table 2 The relationship of CAPM-implied sensitivity and cash flow characteristics with the stock price sensitivity to monetary policy All firms, scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{it-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ Variables (1) CAPM implied sensitivity (2) Cash flow estimated sensitivity (3) Cash flow duration (4) Cash flow variation (5) Book-to-market (6) log market value PolicySurprise 2.04*** 2.08*** 2.36*** 2.08*** 2.36*** 2.31*** (0.49) (0.33) (0.36) (0.33) (0.35) (0.34) PolicySurprise*Characteristic 0.13 0.01 0.48*** 0.29*** –0.14 1.07*** (0.18) (0.07) (0.17) (0.09) (0.14) (0.10) Characteristic 0.02* –0.00 –0.06*** 0.00 0.09*** –0.21*** (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Constant 0.12*** 0.48*** –3.66*** 0.48*** –1.03*** –1.13*** (0.03) (0.18) (0.02) (0.18) (0.23) (0.25) Number of observations 446,653 341,144 288,797 341,117 362,841 445,923 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 Number of stocks 10,986 8,044 7,246 8,043 9,587 11,320 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes All firms, scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{it-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ Variables (1) CAPM implied sensitivity (2) Cash flow estimated sensitivity (3) Cash flow duration (4) Cash flow variation (5) Book-to-market (6) log market value PolicySurprise 2.04*** 2.08*** 2.36*** 2.08*** 2.36*** 2.31*** (0.49) (0.33) (0.36) (0.33) (0.35) (0.34) PolicySurprise*Characteristic 0.13 0.01 0.48*** 0.29*** –0.14 1.07*** (0.18) (0.07) (0.17) (0.09) (0.14) (0.10) Characteristic 0.02* –0.00 –0.06*** 0.00 0.09*** –0.21*** (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Constant 0.12*** 0.48*** –3.66*** 0.48*** –1.03*** –1.13*** (0.03) (0.18) (0.02) (0.18) (0.23) (0.25) Number of observations 446,653 341,144 288,797 341,117 362,841 445,923 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 Number of stocks 10,986 8,044 7,246 8,043 9,587 11,320 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Regressions 2–6 include firm fixed effects. Errors in parentheses are clustered at the firm level. $$^{***}$$ p<.01, $$^{**}$$ p<.05, $$^*$$ p<.1. The dependent variable is the stock returns of publicly listed companies on FOMC announcement dates. PolicySurprise is the unexpected federal funds target rate change on a scheduled FOMC meeting; it is scaled so that positive surprises indicate expansionary shocks. Both returns and PolicySurprise are expressed in percentage points. The coefficient of PolicySurprise refers to the base industry which is nondurables. CAPM-implied sensitivity is the product of the CAPM beta of the stock and the policy sensitivity of CRSP value-weighted returns. The characteristics are defined in Table 1; they are standardized to zero mean and unit variance. The sample includes firms traded in AMEX, NYSE, and NASDAQ. To ensure the liquidity of the stocks, penny stocks are dropped, following Amihud (2002), who defines penny stocks as stocks with a price less than $\$$5. To study other firm characteristics that can play a significant role in explaining cross-sectional differences in the monetary policy sensitivity of stock prices, the rest of this section uses regressions of the form \begin{eqnarray} r_{it} &=&\alpha +\beta \times \textit{PolicySurprise}_{t}+\gamma \times \textit{Characteristic}_{it-1} \label{eq: Characteristic_Regression} \\ &&+\delta \times \textit{Characteristic}_{it-1}\times \textit{PolicySurprise}_{t}+\textit{Controls}+e_{it}, \nonumber \end{eqnarray} (5) where $$r_{it}$$ is the return of stock $$i$$ on FOMC announcement date $$t$$ and $$Characteristic_{it-1}$$ is the most recent value of the firm characteristic. Controls include the industry fixed effects and their interaction with $$PolicySurprise_{t}$$. We are primarily interested in the coefficient $$\delta $$ of $$Characteristic_{it-1}\times PolicySurprise_{t}$$. It is natural to think that characteristics of cash flows may affect the sensitivity of stock prices to monetary policy, although this has often been overlooked in the literature. For example, firms with more interest-sensitive cash flows may have stock prices that are more responsive to monetary policy (Bernanke and Gertler (1995)). Table 2 reports the results for various cash flow characteristics. The first measure of the interest sensitivity of cash flows comes from the regression of cash flow growth on the 3-month Libor rate. As Column 2 shows, this measure does not seem to be a significant source of variation in monetary policy sensitivities because the coefficient $$\delta $$ is both economically and statistically insignificant. The second measure, used in Column 3, is cash flow duration. We expect the value of stocks with longer-duration cash flows to be more sensitive to interest rates because cash flows further in the future are discounted at a higher rate. Indeed, a 1-standard-deviation (5.14 years) increase in cash flow duration generates an additional 0.48-percentage-point increase in the stock price in response to a 1-percentage-point expansionary policy surprise. While this number is reasonably large when compared with the aggregate market reaction of 3.2 percentage points, it is, in absolute terms, much smaller than what we would expect. In particular, this number should be closer to 5.14 than 0.48 because duration is closely related to the derivative of the (log) market value of the assets of the firm with respect to the underlying discount rate. The third measure is the cash flows’ coefficient of variation, a measure of volatility described in Table 1. Cash flow variation can be related to cash flow duration and can capture aspects thereof not perfectly captured by the standard cash flow duration measure. On the one hand, firms with lower volatility may have a lower default likelihood and therefore longer lives and a higher duration of cash flows. On the other hand, a lower volatility may also imply a lower value of the option to delay investment and therefore firms with lower volatility may increase cash flow duration by increasing investment today in exchange for cash flows in the future. Column 4 shows that a 1-standard-deviation increase in cash flow variation generates an additional 0.29-percentage-point increase in the stock price in response to a 1-percentage-point expansionary policy surprise. To summarize, the results thus far indicate that a firm’s industry and cash flow characteristics play a significant role in determining the monetary policy sensitivity of its stock. Of course, the role played by the cash flow characteristics is not completely disentangled from other effects that work through a firm’s capital structure. For example, to the extent that investors find it harder to predict more volatile cash flows or cash flows further in the future, firms with larger cash flow volatility or duration may have trouble accessing external financing. Alternatively, more financially constrained firms may be more conservative in their choice of investments to avoid being caught with insufficient cash flows when needed and hence may choose projects that have lower volatility and duration. This brings us to the next mechanism of monetary policy transmission, perhaps the most famous one, that works through financial frictions. The roots of the interaction between capital structure and monetary policy transmission, the balance sheet channel, can be traced back to Fisher (1933). On the one hand, looser monetary policy can increase the value of the collateral of a credit-rationed firm, which in turn will increase its borrowing capacity, as discussed in Kiyotaki and Moore (1997) and Holmström, and Tirole (1997). Therefore, this firm’s market value may react to monetary policy potentially more strongly than the market value of an unconstrained firm if the value of collateral is sensitive enough to monetary policy. On the other hand, if a firm is constrained due to high bankruptcy costs in a trade-off model (Kraus and Litzenberger (1973)) or due to the information costs in a costly state verification model (Townsend (1979)), it will not be able to borrow as much as an unconstrained firm. Therefore, the fluctuations in the cost of debt caused by monetary policy will affect a constrained firm’s stock price less than the stock price of an unconstrained firm. The appendix illustrates this channel in the context of the widely studied Bernanke, Gertler, and Gilchrist (1999) framework. As these different effects do not necessarily go in the same direction, it is not clear a priori what we should expect to find in the data when we study how financial constraints affect the reaction of stock prices to monetary policy. Because every mechanism that relates capital structure to monetary policy transmission relies on some violation of Modigliani and Miller assumptions, the papers that study the reaction of stock prices to monetary policy focus on financial frictions. However, the previous literature reaches mixed conclusions, reflecting the challenging task of identifying financial frictions. This literature typically identifies smaller firms as financially constrained. Perez-Quiros and Timmermann (2000) find that small firms’ stock prices react more strongly to monetary policy, by using the lagged change in the monetary base as a proxy for monetarypolicy decisions. Lamont, Polk, and Saá-Requejo (2001) recognize that modern monetary policy works through interest rates and hence study the effects of the federal funds rate and the discount window rate, but they find no evidence that the relative performance of constrained firms reflects monetary policy, credit conditions, or business cycles. In contrast, Ehrmann and Fratscher (2004) study S&P 500 firms and find that stock prices of firms that are small or have a high Tobin’s q react more significantly to monetary policy. However, their results do not seem to hold when we consider the universe of CRSP stocks. For example,Column 5 of Table 2 shows that book-to-market values do not seem to play a significant role in policy sensitivity. Column 6 shows that smaller firms seem to be significantly less sensitive to monetary policy than larger firms: a 1-standard-deviation increase in size leads to about a 1-percentage-point increase in the policy sensitivity of stock price. This mixed evidence suggests that we need a more careful study of the role of financial frictions in monetary policy transmission to stock prices. To achieve this goal, I start with modern proxies for financial constraints. When we look at recent financial constraint measures, the results seem to suggest that financially constrained firms are less responsive to monetary policy. To show this, I use a specification similar to regression (5) where $$Characteristic_{it-1}$$ captures whether firm $$i$$ at time $$t$$ is financially constrained based on its most recent value of a given financial constraint measure. These measures come from Kaplan and Zingales (1997), Whited and Wu (2006), and Hadlock and Pierce (2010). The coefficient of interest is again $$\delta $$. The results are reported in the first three columns of Table 3. While the Kaplan and Zingales (KZ) index does not suggest any significant relationship between financial constraints and monetary policy sensitivity, financially constrained firms according to the Hadlock and Pierce (HP) and Whited and Wu (WW) indices respond between 0.6% to 1.1% less to a 1-percentage-point surprise decrease in the policy rate.5 The lack of significance of the Kaplan-Zingales index is not surprising, given that Hadlock and Pierce (2010) developed their index citing the relatively poor performance of the Kaplan-Zingales index in capturing financial constraints.6 In all of these regressions, penny stocks (those with price below $$\$5$$) are excluded to ensure that the stocks in the sample are liquid. Moreover, Columns 4 to 6of Table 3 report the same regressions for stocks with high trading volumes as a further control for liquidity, and the results remain similar. This result is not surprising, given that these FOMC meetings are macroeconomic events scheduled well in advance and therefore their effects should be incorporated to every stock’s price very quickly. Table 3 Financial constraints and stock price sensitivity to monetary policy Scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ All firms High volume firms All firms two-day returns Variables (1) KZ (2) HP (3) WW (4) KZ (5) HP (6) WW (7) KZ (8) HP (9) WW PolicySurprise 2.27*** 2.48*** 2.41*** 2.52*** 2.44*** 2.39*** –0.44 –0.44 0.89*** (0.38) (0.34) (0.35) (0.60) (0.54) (0.54) (0.50) (0.50) (0.12) PolicySurprise*Characteristic –0.01 –0.63*** –1.06*** –0.07 –1.15* –1.34** 0.09 –0.97*** –0.63* (0.25) (0.23) (0.28) (0.41) (0.61) (0.57) (0.05) (0.33) (0.39) Characteristic 0.04* –0.02 0.01 0.10** 0.16*** 0.11** 0.12*** –0.01 0.01 (0.03) (0.03) (0.03) (0.04) (0.06) (0.05) (0.04) (0.04) (0.04) CRSPVW 1.02*** 0.80*** 0.89*** (0.02) (0.01) (0.01) CRSPVW*Characteristic –0.18*** 0.14*** 0.12*** (0.02) (0.02) (0.02) Constant 0.27*** –0.92*** 0.29*** 0.12*** 0.71 0.14*** –0.39*** 1.33*** –0.34*** (0.01) (0.29) (0.01) (0.02) (0.49) (0.01) (0.02) (0.28) (0.02) Number of observations 242,525 446,696 256,528 92,800 151,359 101,626 239,828 437,284 253,754 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.06 0.07 Number of stocks 5,493 11,320 5,546 4,150 7,426 4,214 5,488 11,273 5,541 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Yes Yes Yes Scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ All firms High volume firms All firms two-day returns Variables (1) KZ (2) HP (3) WW (4) KZ (5) HP (6) WW (7) KZ (8) HP (9) WW PolicySurprise 2.27*** 2.48*** 2.41*** 2.52*** 2.44*** 2.39*** –0.44 –0.44 0.89*** (0.38) (0.34) (0.35) (0.60) (0.54) (0.54) (0.50) (0.50) (0.12) PolicySurprise*Characteristic –0.01 –0.63*** –1.06*** –0.07 –1.15* –1.34** 0.09 –0.97*** –0.63* (0.25) (0.23) (0.28) (0.41) (0.61) (0.57) (0.05) (0.33) (0.39) Characteristic 0.04* –0.02 0.01 0.10** 0.16*** 0.11** 0.12*** –0.01 0.01 (0.03) (0.03) (0.03) (0.04) (0.06) (0.05) (0.04) (0.04) (0.04) CRSPVW 1.02*** 0.80*** 0.89*** (0.02) (0.01) (0.01) CRSPVW*Characteristic –0.18*** 0.14*** 0.12*** (0.02) (0.02) (0.02) Constant 0.27*** –0.92*** 0.29*** 0.12*** 0.71 0.14*** –0.39*** 1.33*** –0.34*** (0.01) (0.29) (0.01) (0.02) (0.49) (0.01) (0.02) (0.28) (0.02) Number of observations 242,525 446,696 256,528 92,800 151,359 101,626 239,828 437,284 253,754 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.06 0.07 Number of stocks 5,493 11,320 5,546 4,150 7,426 4,214 5,488 11,273 5,541 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Yes Yes Yes High volume firms refer to those that belong to the highest tercile when firms are ranked according to their trading volume on the day before the FOMC announcement; these firms are more likely to have nonmissing balance sheet information on FOMC dates. All regressions include firm fixed effects. Errors in parentheses are clustered at the firm level. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. The dependent variable is the stock returns of publicly listed companies on FOMC announcement dates. PolicySurprise is the unexpected federal funds target rate change on a scheduled FOMC meeting, as implied by federal funds futures. PolicySurprise is scaled so that positive surprises indicate expansionary shocks. Both returns and PolicySurprise are expressed in percentage points. CRSPVW is the two-day value-weighted returns. Following previous literature, firms are grouped into discrete categories as financially constrained and unconstrained firms. KZ is a dummy variable equal to one if a firm’s Kaplan-Zingales financial constraint measure is above the median in a given year. HP and WW are dummy variables that are similarly defined for Hadlock-Pierce and Whited-Wu financial constraint measure. Further details about variables are in Table 1. The sample includes firms traded in AMEX, NYSE, and NASDAQ. To ensure the liquidity of the stocks, penny stocks are dropped, following Amihud (2002), who defines penny stocks as stocks with a price less than $\$$5. The coefficient of PolicySurprise refers to the base industry which is nondurables. Table 3 Financial constraints and stock price sensitivity to monetary policy Scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ All firms High volume firms All firms two-day returns Variables (1) KZ (2) HP (3) WW (4) KZ (5) HP (6) WW (7) KZ (8) HP (9) WW PolicySurprise 2.27*** 2.48*** 2.41*** 2.52*** 2.44*** 2.39*** –0.44 –0.44 0.89*** (0.38) (0.34) (0.35) (0.60) (0.54) (0.54) (0.50) (0.50) (0.12) PolicySurprise*Characteristic –0.01 –0.63*** –1.06*** –0.07 –1.15* –1.34** 0.09 –0.97*** –0.63* (0.25) (0.23) (0.28) (0.41) (0.61) (0.57) (0.05) (0.33) (0.39) Characteristic 0.04* –0.02 0.01 0.10** 0.16*** 0.11** 0.12*** –0.01 0.01 (0.03) (0.03) (0.03) (0.04) (0.06) (0.05) (0.04) (0.04) (0.04) CRSPVW 1.02*** 0.80*** 0.89*** (0.02) (0.01) (0.01) CRSPVW*Characteristic –0.18*** 0.14*** 0.12*** (0.02) (0.02) (0.02) Constant 0.27*** –0.92*** 0.29*** 0.12*** 0.71 0.14*** –0.39*** 1.33*** –0.34*** (0.01) (0.29) (0.01) (0.02) (0.49) (0.01) (0.02) (0.28) (0.02) Number of observations 242,525 446,696 256,528 92,800 151,359 101,626 239,828 437,284 253,754 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.06 0.07 Number of stocks 5,493 11,320 5,546 4,150 7,426 4,214 5,488 11,273 5,541 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Yes Yes Yes Scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ All firms High volume firms All firms two-day returns Variables (1) KZ (2) HP (3) WW (4) KZ (5) HP (6) WW (7) KZ (8) HP (9) WW PolicySurprise 2.27*** 2.48*** 2.41*** 2.52*** 2.44*** 2.39*** –0.44 –0.44 0.89*** (0.38) (0.34) (0.35) (0.60) (0.54) (0.54) (0.50) (0.50) (0.12) PolicySurprise*Characteristic –0.01 –0.63*** –1.06*** –0.07 –1.15* –1.34** 0.09 –0.97*** –0.63* (0.25) (0.23) (0.28) (0.41) (0.61) (0.57) (0.05) (0.33) (0.39) Characteristic 0.04* –0.02 0.01 0.10** 0.16*** 0.11** 0.12*** –0.01 0.01 (0.03) (0.03) (0.03) (0.04) (0.06) (0.05) (0.04) (0.04) (0.04) CRSPVW 1.02*** 0.80*** 0.89*** (0.02) (0.01) (0.01) CRSPVW*Characteristic –0.18*** 0.14*** 0.12*** (0.02) (0.02) (0.02) Constant 0.27*** –0.92*** 0.29*** 0.12*** 0.71 0.14*** –0.39*** 1.33*** –0.34*** (0.01) (0.29) (0.01) (0.02) (0.49) (0.01) (0.02) (0.28) (0.02) Number of observations 242,525 446,696 256,528 92,800 151,359 101,626 239,828 437,284 253,754 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.06 0.07 Number of stocks 5,493 11,320 5,546 4,150 7,426 4,214 5,488 11,273 5,541 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Yes Yes Yes High volume firms refer to those that belong to the highest tercile when firms are ranked according to their trading volume on the day before the FOMC announcement; these firms are more likely to have nonmissing balance sheet information on FOMC dates. All regressions include firm fixed effects. Errors in parentheses are clustered at the firm level. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. The dependent variable is the stock returns of publicly listed companies on FOMC announcement dates. PolicySurprise is the unexpected federal funds target rate change on a scheduled FOMC meeting, as implied by federal funds futures. PolicySurprise is scaled so that positive surprises indicate expansionary shocks. Both returns and PolicySurprise are expressed in percentage points. CRSPVW is the two-day value-weighted returns. Following previous literature, firms are grouped into discrete categories as financially constrained and unconstrained firms. KZ is a dummy variable equal to one if a firm’s Kaplan-Zingales financial constraint measure is above the median in a given year. HP and WW are dummy variables that are similarly defined for Hadlock-Pierce and Whited-Wu financial constraint measure. Further details about variables are in Table 1. The sample includes firms traded in AMEX, NYSE, and NASDAQ. To ensure the liquidity of the stocks, penny stocks are dropped, following Amihud (2002), who defines penny stocks as stocks with a price less than $\$$5. The coefficient of PolicySurprise refers to the base industry which is nondurables. As a final robustness check, the last three columns present results with 2-day returns and also include 2-day CRSP value-weighted returns interacted with proxies of financial constraints to control for constrained and unconstrained firms’ potentially different sensitivities to other market-related news. Overall, the results are similar, although the coefficient of the policy surprise is smaller due to inclusion of the CRSP value-weighted returns and the coefficient of policy surprise interacted with friction proxies have larger standard errors due to a less tight event window, as expected. Moreover, the post-FOMC blackout period, during which Federal Reserve employees are not allowed to comment on current monetary policy, usually lasts two days, making the inference based on a wider event window not as reliable.7 This explains the preference of earlier papers to study daily or intraday returns for better identification of the effect of monetary policy, a preference the rest of the paper will follow.8 These results seem to be more in line with the argument that financial constraints can reduce the responsiveness of stock prices to monetary policy surprises. Incidentally, the median book leverage (Compustat (DLC+DLTT)/AT) of financially constrained firms according to the HP and WW indices is much smaller than the median book leverage of unconstrained firms, about 0.06 vs 0.22, which seems consistent with the effect of leverage being muted for the constrained firms.9 Financial frictions underlying the balance sheet channel are usually a manifestation of information frictions. Indeed, Gertler and Gilchrist (1994), a paper widely cited in previous literature, defends the use of firm size as a financial constraint measure by arguing that “the informational frictions that add to the costs of external finance apply mainly to [$$\ldots$$] smaller firms.” In the rest of the paper, I establish the special role of information frictions and clarify the channels through which information frictions operate. I start with suggestive evidence from commonly used measures of information frictions, for example, in Sloan (1996) and Sufi (2007), including whether the firm is unrated, has high R&D expenses, or has positive accruals and hence is more likely to engage in earnings management. I use the same regression specification as equation (5), where a firm is considered financially constrained based on the most recent value of the information friction measure for firm $$i$$ at time $$t$$. Table 4 presents the results when firms are grouped according to these measures. The results are largely consistent with the results from the Hadlock and Pierce and Whited and Wu indices. Unrated firms respond by about 1.30 percentage points less than others to a 1-percentage-point expansionary policy surprise. Similar to the case of the HP and WW indices, the median leverage of unrated firms is lower than that of rated firms, about 0.08 versus 0.31. While there does not seem to be a significant difference between firms with high and low R&D expenses, firms with positive accruals seem to have lower monetary policy sensitivity than other firms, consistent with the results that come from ratings availability. These results are robust and even become somewhat stronger in the case of R&D spending and accruals when I use the subsample of stocks with high trading volume, as shown in Columns 4–6. For completeness, the last three columnsof Table 4 present results from 2-day returns as I did for the financial constraint proxies in Table 3. These results confirm the conclusions from the last three columns of Table 3: the coefficients of policy surprise are smaller due to inclusion of the CRSP value-weighted returns and standard errors are larger for the coefficient of policy surprise interacted with friction proxies due to the less tight event window, which makes the estimates from tighter event windows preferable, as in previous literature. Table 4 Information frictions and stock price sensitivity to monetary policy Scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ All firms High volume firms All firms two-day returns Variables (1) Unrated (2) High R&D (3) Positive accruals (4) Unrated (5) High R&D (6) Positive accruals (7) Unrated (8) High R&D (9) Positive accruals PolicySurprise 3.17*** 2.13*** 2.41*** 2.74*** 1.62* 2.74*** 0.57 0.66*** 0.09 (0.35) (0.74) (0.35) (0.55) (0.83) (0.58) (0.45) (0.17) (0.47) PolicySurprise*Characteristic –1.30*** –0.09 –0.55** –1.27*** –0.63 –1.33*** –0.70*** 0.02 –0.64* (0.17) (0.37) (0.28) (0.36) (0.61) (0.51) (0.24) (0.53) (0.37) Characteristic –0.00 0.05 –0.07*** 0.09** 0.13* –0.07* 0.08** –0.18*** –0.08*** (0.02) (0.05) (0.02) (0.04) (0.07) (0.03) (0.03) (0.06) (0.03) CRSPVW 0.83*** 0.81*** 0.88*** (0.01) (0.01) (0.01) CRSPVW*Characteristic 0.03* 0.02 0.09*** (0.02) (0.53) (0.02) Constant –1.13*** 0.28*** 1.99*** –0.39 0.26*** 0.14*** 2.56*** –0.31*** 1.77*** (0.25) (0.03) (0.68) (0.43) (0.04) (0.01) (0.37) (0.04) (0.01) Number of observations 447,173 211,592 316,760 151,438 85,791 120,430 437,713 209,308 312,831 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 0.06 0.07 0.06 Number of stocks 11,348 5,670 7,958 7,435 4,148 5,680 11,297 5,658 7,933 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Yes Yes Yes Scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ All firms High volume firms All firms two-day returns Variables (1) Unrated (2) High R&D (3) Positive accruals (4) Unrated (5) High R&D (6) Positive accruals (7) Unrated (8) High R&D (9) Positive accruals PolicySurprise 3.17*** 2.13*** 2.41*** 2.74*** 1.62* 2.74*** 0.57 0.66*** 0.09 (0.35) (0.74) (0.35) (0.55) (0.83) (0.58) (0.45) (0.17) (0.47) PolicySurprise*Characteristic –1.30*** –0.09 –0.55** –1.27*** –0.63 –1.33*** –0.70*** 0.02 –0.64* (0.17) (0.37) (0.28) (0.36) (0.61) (0.51) (0.24) (0.53) (0.37) Characteristic –0.00 0.05 –0.07*** 0.09** 0.13* –0.07* 0.08** –0.18*** –0.08*** (0.02) (0.05) (0.02) (0.04) (0.07) (0.03) (0.03) (0.06) (0.03) CRSPVW 0.83*** 0.81*** 0.88*** (0.01) (0.01) (0.01) CRSPVW*Characteristic 0.03* 0.02 0.09*** (0.02) (0.53) (0.02) Constant –1.13*** 0.28*** 1.99*** –0.39 0.26*** 0.14*** 2.56*** –0.31*** 1.77*** (0.25) (0.03) (0.68) (0.43) (0.04) (0.01) (0.37) (0.04) (0.01) Number of observations 447,173 211,592 316,760 151,438 85,791 120,430 437,713 209,308 312,831 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 0.06 0.07 0.06 Number of stocks 11,348 5,670 7,958 7,435 4,148 5,680 11,297 5,658 7,933 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Yes Yes Yes High Volume Firms refer those that belong to the highest tercile when firms are ranked according to their trading volume on the day before the FOMC announcement; these firms are more likely to have nonmissing balance sheet information on FOMC dates. All regressions include firm fixed effects. Errors in parentheses are clustered at the firm level. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. The dependent variable is the stock returns of publicly listed companies on FOMC announcement dates. PolicySurprise is the unexpected federal funds target rate change on a scheduled FOMC meeting, as implied by federal funds futures. PolicySurprise is scaled so that positive surprises indicate expansionary shocks. Both returns and PolicySurprise are expressed in percentage points. CRSPVW is the two-day value-weighted returns. Following previous literature, firms are grouped into discrete categories as financially constrained and unconstrained firms. Unrated is a dummy variable equal to one if a firm does not have a long-term credit rating. High R&D is a dummy variable equal to one if a firm’s R&D spending relative to assets (XRD/AT) is above the median in a given year. Positive Accruals is a dummy variable equal to one if the firms has positive accruals. Further details about variables are in Table 1. The sample includes firms traded in AMEX, NYSE, and NASDAQ. To ensure the liquidity of the stocks, penny stocks are dropped, following Amihud (2002), who defines penny stocks as stocks with a price less than $\$$5. The coefficient of PolicySurprise refers to the base industry which is nondurables. Table 4 Information frictions and stock price sensitivity to monetary policy Scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ All firms High volume firms All firms two-day returns Variables (1) Unrated (2) High R&D (3) Positive accruals (4) Unrated (5) High R&D (6) Positive accruals (7) Unrated (8) High R&D (9) Positive accruals PolicySurprise 3.17*** 2.13*** 2.41*** 2.74*** 1.62* 2.74*** 0.57 0.66*** 0.09 (0.35) (0.74) (0.35) (0.55) (0.83) (0.58) (0.45) (0.17) (0.47) PolicySurprise*Characteristic –1.30*** –0.09 –0.55** –1.27*** –0.63 –1.33*** –0.70*** 0.02 –0.64* (0.17) (0.37) (0.28) (0.36) (0.61) (0.51) (0.24) (0.53) (0.37) Characteristic –0.00 0.05 –0.07*** 0.09** 0.13* –0.07* 0.08** –0.18*** –0.08*** (0.02) (0.05) (0.02) (0.04) (0.07) (0.03) (0.03) (0.06) (0.03) CRSPVW 0.83*** 0.81*** 0.88*** (0.01) (0.01) (0.01) CRSPVW*Characteristic 0.03* 0.02 0.09*** (0.02) (0.53) (0.02) Constant –1.13*** 0.28*** 1.99*** –0.39 0.26*** 0.14*** 2.56*** –0.31*** 1.77*** (0.25) (0.03) (0.68) (0.43) (0.04) (0.01) (0.37) (0.04) (0.01) Number of observations 447,173 211,592 316,760 151,438 85,791 120,430 437,713 209,308 312,831 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 0.06 0.07 0.06 Number of stocks 11,348 5,670 7,958 7,435 4,148 5,680 11,297 5,658 7,933 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Yes Yes Yes Scheduled FOMC announcement dates, 1994–2008 $$\textit{return}_{\textit{it}} = \alpha + \beta * {\textit{PolicySurprise}}_t + \gamma * {\textit{Characteristic}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Characteristic}}_{\textit{it}-1} + e_{\textit{it}}$$ All firms High volume firms All firms two-day returns Variables (1) Unrated (2) High R&D (3) Positive accruals (4) Unrated (5) High R&D (6) Positive accruals (7) Unrated (8) High R&D (9) Positive accruals PolicySurprise 3.17*** 2.13*** 2.41*** 2.74*** 1.62* 2.74*** 0.57 0.66*** 0.09 (0.35) (0.74) (0.35) (0.55) (0.83) (0.58) (0.45) (0.17) (0.47) PolicySurprise*Characteristic –1.30*** –0.09 –0.55** –1.27*** –0.63 –1.33*** –0.70*** 0.02 –0.64* (0.17) (0.37) (0.28) (0.36) (0.61) (0.51) (0.24) (0.53) (0.37) Characteristic –0.00 0.05 –0.07*** 0.09** 0.13* –0.07* 0.08** –0.18*** –0.08*** (0.02) (0.05) (0.02) (0.04) (0.07) (0.03) (0.03) (0.06) (0.03) CRSPVW 0.83*** 0.81*** 0.88*** (0.01) (0.01) (0.01) CRSPVW*Characteristic 0.03* 0.02 0.09*** (0.02) (0.53) (0.02) Constant –1.13*** 0.28*** 1.99*** –0.39 0.26*** 0.14*** 2.56*** –0.31*** 1.77*** (0.25) (0.03) (0.68) (0.43) (0.04) (0.01) (0.37) (0.04) (0.01) Number of observations 447,173 211,592 316,760 151,438 85,791 120,430 437,713 209,308 312,831 R-squared 0.00 0.00 0.00 0.00 0.00 0.00 0.06 0.07 0.06 Number of stocks 11,348 5,670 7,958 7,435 4,148 5,680 11,297 5,658 7,933 Industry dummy*PolicySurprise Yes Yes Yes Yes Yes Yes Yes Yes Yes High Volume Firms refer those that belong to the highest tercile when firms are ranked according to their trading volume on the day before the FOMC announcement; these firms are more likely to have nonmissing balance sheet information on FOMC dates. All regressions include firm fixed effects. Errors in parentheses are clustered at the firm level. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. The dependent variable is the stock returns of publicly listed companies on FOMC announcement dates. PolicySurprise is the unexpected federal funds target rate change on a scheduled FOMC meeting, as implied by federal funds futures. PolicySurprise is scaled so that positive surprises indicate expansionary shocks. Both returns and PolicySurprise are expressed in percentage points. CRSPVW is the two-day value-weighted returns. Following previous literature, firms are grouped into discrete categories as financially constrained and unconstrained firms. Unrated is a dummy variable equal to one if a firm does not have a long-term credit rating. High R&D is a dummy variable equal to one if a firm’s R&D spending relative to assets (XRD/AT) is above the median in a given year. Positive Accruals is a dummy variable equal to one if the firms has positive accruals. Further details about variables are in Table 1. The sample includes firms traded in AMEX, NYSE, and NASDAQ. To ensure the liquidity of the stocks, penny stocks are dropped, following Amihud (2002), who defines penny stocks as stocks with a price less than $\$$5. The coefficient of PolicySurprise refers to the base industry which is nondurables. The plethora of financial constraint indices use different firm characteristics, which blurs the precise mechanism through which the financial frictions operate. The identification of the role of financial frictions is further complicated by the endogeneity of the financial constraint proxies. Farre-Mensa and Ljungqvist (2015) find that firms typically classified as constrained by these proxies do not in fact behave as if they were constrained, and the financial constraint measures may instead reflect differences in the growth and financing policies of firms at different stages of their lifecycles. These differences can be a source of omitted variable bias. For example, suppose a high-WW firm is not really financially constrained but is actually a firm with large growth options. Then, the estimatesin Table 3 are biased toward zero because growth options have higher cash flow duration and higher implicit leverage (Berk, Green, and Naik (2004)), which can make a firm’s stock price more responsive to monetary policy. While it is encouraging that the results from information friction proxies in Table 4 point in the same direction as financial constraint proxies in Table 3, there is no guarantee that the information friction proxies are totally immune to the endogeneity problems either. Therefore, the next section follows with a natural experiment to better identify the role of information frictions and the channels through which they operate. 2. Evidence from a Natural Experiment: The Enron Scandal I use the Enron accounting scandal of 2001 and the resultant demise of Enron’s auditing firm, Arthur Andersen, as an exogenous large shock to the information frictions of other Arthur Andersen clients relative to the clients of other auditors. Because information frictions are greatly reduced by the independent auditing of firms’ financial statements, a sudden loss in the credibility of a major auditing firm is an important and long-lasting shock to its clients. Arthur Andersen was one of the “Big Five” accounting firms, including PricewaterhouseCoopers, Deloitte Touche Tohmatsu, Ernst & Young, and KPMG, providing auditing, tax, and consulting services to large corporations. Being one of the “Big Five,” it has enjoyed a positive perception in terms of its auditing quality not only in public but also in academic media, for example, Chaney, Jeter, and Shivakumar (2004), and Chang, Dasgupta, and Hilary (2009). However, this perception came to an abrupt end in summer and fall of 2001 due to accounting problems in Enron, a large energy corporation that has been a client of Andersen for 16 years up until Enron’s 2001 bankruptcy. Andersen is thought to have done questionable accounting for Enron, which hid millions of dollars of debt from the public. The Enron scandal hit markets as a surprise because Enron was a market favorite before summer 2001. Many industry experts, such as Campbell and Hulme (2001), were praising Enron’s business model even when Enron’s accounting quality were first put under public scrutiny (McLean (2001)). Still, the doubts about Enron intensified and reached new heights on October 16, 2001, when Enron posted huge losses in shareholder’s equity. On October 22, 2001, Enron announced that the Securities and Exchange Commission (SEC) had requested information regarding certain related-party transactions. Enron’s stock price fell from $\$$33.84 on October 16, 2001, to $\$$8.63 in the days following the restatement announcement on November 8, 2001. This has sealed Enron’s and Arthur Andersen’s fate: Enron filed for bankruptcy on December 2, 2001, Andersen was charged with obstruction of justice for destroying the documents related to its audit of Enron and was convicted on June 15, 2002. Although the conviction was later reversed by the Supreme Court in 2005, Arthur Andersen’s reputation has remained irreparably tarnished, which cast a cloud on the financial statements of its clients. There are three implicit assumptions in my identification approach. The first is that a firm’s choice of auditor did not have any sizable effect on the market valuation of the firm before the scandal. In particular, prior to the Enron scandal, the perceived probability that an Arthur Andersen client would engage in fraud was not different from the probability that clients of other auditing firms would do so. Consistent with this assumption, Eisenberg and Macey (2004) and Agrawal, and Chadha (2005) find that Arthur Andersen clients did not get involved with more financial restatements than other firms in the years preceding the Enron scandal. Similarly, Dyck, Morse, and Zingales (2014) find that Arthur Andersen clients were not more likely to engage in earnings manipulation than other firms, and in their study of the economic cost of fraud, they make the same identification assumption used in this paper.10 The second assumption relies on the contagion effects of possible accounting fraud from sharing the same auditing firm: the perceived reliability of financial statements by other Arthur Andersen clients decreased relative to clients of other auditors. Consistent with this assumption, Dyck, Morse, and Zingales (2014) find that the probability of fraud detection increased dramatically for former Arthur Andersen clients once they switched to another auditor after the Enron scandal. Moreover, Gleason, Jenkins, and Johnson (2008) use stock returns to provide evidence for the contagion effect that accounting restatements have on the credibility of financial statements issued by nonrestating firms. They argue that the contagion effect is unrelated to changes in analysts’ earnings-per-share forecasts, and hence to the economic prospects of the firms, and that it is stronger for the firms with low accounting quality (high accruals) that share the same external auditor. The final assumption is that the reliability of financial statements, and hence auditing, affects firms’ financing outcomes. Given that the main purpose of financial accounting is to prepare financial reports that provide information about a firm’s performance to external parties, thereby reducing the information frictions faced by investors, this assumption may seem trivial. However, some lenders, such as banks, may have superior information due to their relationship with a firm and hence may not need the auditors’ opinion. Still, banks need to rely on financial statements because bank loans have covenants based on financial statement variables. Therefore, the absence of reliable financial statements may strain lender-firm relationships, increasing the cost of borrowing for the firms. Consistent with this observation, Graham, Li, and Qiu (2008) find that financial restatements lead to increased bank loan rates and tighter covenants for the restating firms. Combined with the finding of Dyck, Morse, and Zingales (2014) that the probability of financial restatements has increased for Arthur Andersen clients after the Enron scandal, we can conclude that investors likely expected that Arthur Andersen clients would face greater financial constraints due to the scandal. One could argue that the firms might switch their auditors immediately upon learning the news of the scandal, so that new auditors could calm investors’ concerns, thereby alleviating financial constraints. However, the data say that firms could not switch auditors immediately: although the news about Enron’s fraud has progressed steadily during 2001, almost all the firms remained with Arthur Andersen until the end of 2001 and even for a while thereafter (at least until the end of March 2002, as firms file their statements three months after the end of their fiscal year), most likely because of contractual obligations. Moreover, “cleaning house,” as Dyck, Morse, and Zingales (2014) put it, takes time even after the switch because it is not straightforward for the new auditors to vet the old financial statements. Consistent with this discussion, Figure 2 shows that while Arthur Andersen clients’ leverage moved in tandem with the leverage of other firms before the Enron scandal, their leverage dropped 20% more than the leverage of other firms in the years following the scandal. The effect on leverage is delayed because existing debt does not mature immediately, generating persistence of capital structure. This effect peaks in year 2005, which coincides with the Supreme Court’s overruling of Arthur Andersen’s conviction on May 31, 2005. It takes about four years for the difference in leverage between Arthur Andersen clients and other firms to start disappearing.11 Figure 2 View largeDownload slide Difference in leverage of Arthur Andersen clients relative to other firms The dots represent the point estimates of the difference between the log(Leverage) of Arthur Andersen clients and that of other firms, after controlling for firm and year fixed effects and other control variables (market-to-book assets, log(sales), profitability, tangibility, and industry median leverage) commonly used in the corporate finance literature. See regression (3) in Table 10 for details. The bars represent 95% confidence intervals. Figure 2 View largeDownload slide Difference in leverage of Arthur Andersen clients relative to other firms The dots represent the point estimates of the difference between the log(Leverage) of Arthur Andersen clients and that of other firms, after controlling for firm and year fixed effects and other control variables (market-to-book assets, log(sales), profitability, tangibility, and industry median leverage) commonly used in the corporate finance literature. See regression (3) in Table 10 for details. The bars represent 95% confidence intervals. 2.1 Auditor and balance sheet data The auditor information and balance sheet data come from Compustat annual files to reduce the number of missing observations and to limit potential problems with the seasonality in quarterly data items (Leary and Roberts (2010)). To ensure the liquidity of the stocks, penny stocks, stocks with a price less than $\$$5, are dropped (Amihud (2002)) and to ensure that the informativeness of financial statements is similar across companies, the sample is restricted to firms having December as their fiscal year-end (Polk and Sapienza (2009)). Finally, the sample excludes firms that have undergone a significant merger or acquisition in 2001 (Kashyap, Lamont, and Stein (1994)). These firms are indicated in the footnote of the Compustat sales item (SALE_FN) (Anantharaman and Lee (2014)). The control variables are the usual suspects, taken from the cross-sectional asset pricing literature. Market Leverage is calculated by dividing the book value of debt by the sum of the book value of debt and the market value of common equity. The market value of common equity is calculated as the stock price times shares outstanding from CRSP as of December 31, 2000. The book value of debt is calculated as total assets minus book equity, where book equity is equal to the sum of common equity and deferred taxes (Compustat items CEQ and TXDITC, respectively), like in Fama and French (1992). Book-to-Market is the book value of equity divided by the market value of equity. Assets are total assets (Compustat Item AT).12 Profitability is operating income (Compustat Item OIBDP) divided by total assets. Finally, the CAPM beta is calculated using daily returns between the current FOMC announcement date and the prior FOMC announcement date.13 Table 5 provides key statistics for the characteristics of the treatment group (Andersen’s clients) and the control group (other firms). The market leverage, book-to-market ratio, assets, and profitability across these groups are remarkably similar and the minor differences are not statistically significant. When we look at common proxies for financial frictions, they are also very similar across the two groups. The differences are statistically insignificant, except for the KZ index, R&D spending, and accruals. However, as we have seen in Section 1, the KZ index is not a strong predictor of policy sensitivity and the differences in R&D spending and accruals between the two groups are economically very small (less than a quarter of a standard deviation). Nevertheless, my robustness checks also analyze firms with high R&D and accruals separately, giving further assurance that these minor differences do not change the main conclusions. Finally, the industry distributions also seem to be similar across the two groups according to the Kolmogorov-Smirnov test ($$p=0.24$$). Overall, the characteristics of Arthur Andersen clients and other firms seem to be reasonably close to each other, alleviating any concern regarding confounding effectsunrelated to the auditors. Table 5 Characteristics for the treatment group (Andersen clients) and the control group (other firms) Andersen clients Other firms Variable # firms Mean Median SD # firms Mean Median SD P(Diff) Market leverage 253 0.39 0.37 0.27 1,038 0.35 0.33 0.26 0.14 Book-to-market 253 0.68 0.49 0.69 1,038 0.63 0.48 0.61 0.48 log(Total assets) 269 6.65 6.40 1.79 1,119 6.60 6.42 1.96 0.57 Profitability 267 0.11 0.12 0.15 1,108 0.09 0.12 0.19 0.12 CAPM beta 269 0.85 0.66 0.84 1,118 0.92 0.72 0.88 0.15 WW 182 –0.29 –0.28 0.21 706 –0.28 –0.28 0.12 0.70 KZ 170 –7.90 –0.23 36.26 666 –7.43 –0.83 28.05 0.09 HP 269 –3.50 –3.37 0.66 1,119 –3.44 –3.33 0.68 0.43 Unrated 269 0.57 1 0.50 1,119 0.61 1 0.49 0.74 R&D spending 119 0.06 0.03 0.11 592 0.09 0.05 0.13 0.06 Accruals 226 –0.05 –0.05 0.09 855 –0.04 –0.04 0.10 0.05 Andersen clients Other firms Variable # firms Mean Median SD # firms Mean Median SD P(Diff) Market leverage 253 0.39 0.37 0.27 1,038 0.35 0.33 0.26 0.14 Book-to-market 253 0.68 0.49 0.69 1,038 0.63 0.48 0.61 0.48 log(Total assets) 269 6.65 6.40 1.79 1,119 6.60 6.42 1.96 0.57 Profitability 267 0.11 0.12 0.15 1,108 0.09 0.12 0.19 0.12 CAPM beta 269 0.85 0.66 0.84 1,118 0.92 0.72 0.88 0.15 WW 182 –0.29 –0.28 0.21 706 –0.28 –0.28 0.12 0.70 KZ 170 –7.90 –0.23 36.26 666 –7.43 –0.83 28.05 0.09 HP 269 –3.50 –3.37 0.66 1,119 –3.44 –3.33 0.68 0.43 Unrated 269 0.57 1 0.50 1,119 0.61 1 0.49 0.74 R&D spending 119 0.06 0.03 0.11 592 0.09 0.05 0.13 0.06 Accruals 226 –0.05 –0.05 0.09 855 –0.04 –0.04 0.10 0.05 Market leverage is calculated by dividing the book value of debt by the sum of the book value of debt and the market value of common equity. The market value of common equity is the stock price times shares outstanding from CRSP as of December 31, 2000. The book value of debt is total assets minus book equity, where book equity is equal to the sum of common equity and deferred taxes (Compustat items CEQ and TXDITC, respectively). Book-to-market is the book value of equity divided by the market value of equity. Assets is total assets (Compustat item AT). Profitability is operating income (Compustat item OIBDP) divided by total assets. All balance sheet items are from Compustat 2000 annual files. CAPM neta is calculated using the daily returns since the prior FOMC meeting. Other variable descriptions, WW to accruals, are in Table 1. P(Diff) is the p-value of the Kolmogorov-Smirnov test of whether the samples have similar distributions. This test is robust to nonnormality and monotonic transformation of the variables. Summary statistics are given for those observations for which the value of the variable is nonmissing and the firm has a nonmissing industry code. For example, the HP index uses only firm size and age, which are available for most firms, and therefore it has more observations than the WW index which has six terms that use seven Compustat variables, some of which can be missing for a given firm. Table 5 Characteristics for the treatment group (Andersen clients) and the control group (other firms) Andersen clients Other firms Variable # firms Mean Median SD # firms Mean Median SD P(Diff) Market leverage 253 0.39 0.37 0.27 1,038 0.35 0.33 0.26 0.14 Book-to-market 253 0.68 0.49 0.69 1,038 0.63 0.48 0.61 0.48 log(Total assets) 269 6.65 6.40 1.79 1,119 6.60 6.42 1.96 0.57 Profitability 267 0.11 0.12 0.15 1,108 0.09 0.12 0.19 0.12 CAPM beta 269 0.85 0.66 0.84 1,118 0.92 0.72 0.88 0.15 WW 182 –0.29 –0.28 0.21 706 –0.28 –0.28 0.12 0.70 KZ 170 –7.90 –0.23 36.26 666 –7.43 –0.83 28.05 0.09 HP 269 –3.50 –3.37 0.66 1,119 –3.44 –3.33 0.68 0.43 Unrated 269 0.57 1 0.50 1,119 0.61 1 0.49 0.74 R&D spending 119 0.06 0.03 0.11 592 0.09 0.05 0.13 0.06 Accruals 226 –0.05 –0.05 0.09 855 –0.04 –0.04 0.10 0.05 Andersen clients Other firms Variable # firms Mean Median SD # firms Mean Median SD P(Diff) Market leverage 253 0.39 0.37 0.27 1,038 0.35 0.33 0.26 0.14 Book-to-market 253 0.68 0.49 0.69 1,038 0.63 0.48 0.61 0.48 log(Total assets) 269 6.65 6.40 1.79 1,119 6.60 6.42 1.96 0.57 Profitability 267 0.11 0.12 0.15 1,108 0.09 0.12 0.19 0.12 CAPM beta 269 0.85 0.66 0.84 1,118 0.92 0.72 0.88 0.15 WW 182 –0.29 –0.28 0.21 706 –0.28 –0.28 0.12 0.70 KZ 170 –7.90 –0.23 36.26 666 –7.43 –0.83 28.05 0.09 HP 269 –3.50 –3.37 0.66 1,119 –3.44 –3.33 0.68 0.43 Unrated 269 0.57 1 0.50 1,119 0.61 1 0.49 0.74 R&D spending 119 0.06 0.03 0.11 592 0.09 0.05 0.13 0.06 Accruals 226 –0.05 –0.05 0.09 855 –0.04 –0.04 0.10 0.05 Market leverage is calculated by dividing the book value of debt by the sum of the book value of debt and the market value of common equity. The market value of common equity is the stock price times shares outstanding from CRSP as of December 31, 2000. The book value of debt is total assets minus book equity, where book equity is equal to the sum of common equity and deferred taxes (Compustat items CEQ and TXDITC, respectively). Book-to-market is the book value of equity divided by the market value of equity. Assets is total assets (Compustat item AT). Profitability is operating income (Compustat item OIBDP) divided by total assets. All balance sheet items are from Compustat 2000 annual files. CAPM neta is calculated using the daily returns since the prior FOMC meeting. Other variable descriptions, WW to accruals, are in Table 1. P(Diff) is the p-value of the Kolmogorov-Smirnov test of whether the samples have similar distributions. This test is robust to nonnormality and monotonic transformation of the variables. Summary statistics are given for those observations for which the value of the variable is nonmissing and the firm has a nonmissing industry code. For example, the HP index uses only firm size and age, which are available for most firms, and therefore it has more observations than the WW index which has six terms that use seven Compustat variables, some of which can be missing for a given firm. 2.2 Empirical strategy and results The Enron scandal was not a sudden event, but rather a scandal that unfolded over the course of 2001. This brings up several challenges. First, while the firms’ capital structure evolves slowly in response to the scandal, stock prices are forward looking and incorporate any new information immediately. Therefore, the wave of accounting scandals starting early 2002 will affect inference, as they involve other auditing firms. This suggests that one should pay more attention to the monetary policy announcements during 2001. However, there is no perfectly reliable way to figure out the evolution of investors’ beliefs about the probability of an accounting scandal over the course of 2001. As a result, it is necessary to use an FOMC announcement date with a sizable monetary policy surprise that occurred late enough in 2001 to incorporate the full effect of the scandal. In addition, unscheduled monetary policy announcements must be omitted to avoid the effect of timing shocks that would reduce the exogeneity of the measured policy surprise. These criteria lead to the choice of the scheduled FOMC announcement on November 6, 2001, because this date includes a sizable monetary policy surprise ($$-10$$ bp) for the $$50$$-bp reduction in the federal funds target rate on that date, and it is very close to November 8, 2001, when Enron filed the 8-K report announcing that it would restate financial documents dating from 1997 through thesecond quarter of 2001.14 As a comparable “before treatment” date, May 15, 2001, is chosen, because the FOMC announcement on this date shares similar characteristics with the FOMC announcement made on November 6, 2001. In particular, the change in the federal funds target rate on both dates was $$-50$$ bp, both were scheduled announcements, and each had a similar size for the monetary policy surprise component, $$-8$$ bp versus $$-10$$ bp.15 Moreover, because both surprises were negative, any estimated difference between the stock price reactions of Arthur Andersen clients and other firms cannot be attributed to asymmetric effects of expansionary and contractionary policy shocks. The other monetary policy actions in early 2001 were either unscheduled, which would introduce timing shocks and violate the exogeneity of the monetary policy surprise, or had zero or positive surprises. The strong policy actions on both dates help significantly in identifying the effect of the policy. Using these two FOMC dates, the first analysis employs a difference-in-differences approach, \begin{eqnarray} \textit{return}_{\textit{it}} &=& \beta _{0}+\beta _{1}\text{AAClient}_{i}+\beta _{2}\text{After} _{t}+\tilde{\delta}\text{AAClient}_{i}\ast \text{After}_{t} \nonumber \\ &&+\text{controls}+e_{it}\text{,} \label{eq: DID_AA} \end{eqnarray} (6) where $$\tilde{\delta}$$ is the parameter of interest. AAClient is a dummy variable equal to one if the firm’s financial statements for the year 2000 were audited by Arthur Andersen, and zero otherwise. After is a dummy variable that is equal to one for the observations on November 6, 2001, and zero for observations on May 15, 2001. This regression includes only firms that have stock return data on both FOMC announcement dates. This approach implicitly controls for firm-specific fixed effects on returns because, in a balanced panel of two dates, a difference-in-differences regression and a fixed effects panel regression provide the same coefficient estimates. Control variables include industry fixed effects and their interaction with After, and other firm characteristics as detailed in Section 2.1. The negative relationship between information frictions and policy sensitivity demonstrated in Section 1 implies that the stock prices of Arthur Andersen clients should have reacted relatively less positively to the expansionary monetary policy shock on November 6 as their information frictions have intensified relative to other firms ($$\tilde{\delta}<0$$). Column 1 in Table 6 presents the results of this regression without using any additional control variables. The first line tells us that following the Enron scandal, the sensitivity of the stock prices of Arthur Andersen clients to a 10-bp reduction in the federal funds target rate decreased by 67 bp relative to that of other auditors’ clients. This effect is very large, considering that, on average, a 10 basis point surprise decrease in the federal funds target rate leads to about a 100 basis point increase in stock prices during this time period.16 Column 2 presents the results of the same regression after including control variables. These controls do not seem to affect the average returns on these two dates, with the exception of the CAPM beta that is positively related with returns. More importantly, the coefficient of interest, that on AAClient*After, practically stays the same.17 Table 6 The ENRON scandal’s effect on stock price sensitivity to monetary policy Arthur Andersen Clients versus Other Firms: May 15, 2001 versus November 6, 2001 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 + {\textit{AAClient}}_i + \beta_2 * {\textit{After}}_t + \tilde{\delta} * {\textit{AAClient}}_i + {\textit{After}}_t + {\textit{controls}} + {e}_{\textit{it}}$$ Variables (1) OLS no controls (2) OLS with controls (3) BIG 5 (4) IV AA1995 (5) IV AA9899 (6) INTRADAY (7) RATED (8) UNRATED (9) RATED 2-day (10) UNRATED 2-day AAClient*After –0.665** –0.630** –0.720** –0.519* –0.768** –0.555*** 0.067 –1.194*** 0.315 –0.781 (0.291) (0.315) (0.312) (0.285) (0.340) (0.215) $$\underbrace {\matrix{{\left( {0.405} \right)} & {\left( {0.453} \right)} \cr} }_{ - {{1.261}^{ * * }}}$$ $$\underbrace {\matrix{{\left( {0.574} \right)} & {\left( {0.781} \right)} \cr } }_{ - 1.096}$$ (0.608) (0.965) AAClient 0.271 0.220 0.259 0.286 0.331 0.455*** –0.110 0.500 –0.241 0.095 (0.190) (0.212) (0.204) (0.199) (0.236) (0.135) (0.249) (0.312) (0.390) (0.585) After 0.332** (0.139) log(Asset) 0.075 0.083 –0.003 0.023 0.116** –0.169 0.259 0.007 0.672** (0.083) (0.080) (0.084) (0.084) (0.052) (0.112) (0.165) (0.152) (0.313) Book-to-market 0.085 0.106 –0.010 0.084 0.051 0.117 0.042 0.083 –0.208 (0.096) (0.095) (0.106) (0.102) (0.076) (0.135) (0.134) (0.198) (0.235) Market leverage –0.104 –0.124 0.034 –0.051 –0.117* –0.182 –0.156 –0.310 –0.115 (0.105) (0.103) (0.122) (0.109) (0.068) (0.189) (0.147) (0.202) (0.268) Profitability 0.066 0.027 0.161 0.217** –0.035 –0.142 0.027 –0.228 –0.013 (0.095) (0.095) (0.140) (0.096) (0.058) (0.285) (0.106) (0.274) (0.184) CAPM beta 0.320*** 0.313*** 0.292* 0.353*** 0.343*** 0.333 0.242* 0.013 0.556** (0.115) (0.117) (0.177) (0.135) (0.077) (0.221) (0.140) (0.300) (0.222) Constant 0.685*** (0.085) Number of obs. 2,776 2,554 2,452 1,725 2,199 1,902 970 1,584 968 1,553 R-squared 0.003 0.027 0.028 0.030 0.031 0.121 0.058 0.029 0.061 0.044 Arthur Andersen Clients versus Other Firms: May 15, 2001 versus November 6, 2001 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 + {\textit{AAClient}}_i + \beta_2 * {\textit{After}}_t + \tilde{\delta} * {\textit{AAClient}}_i + {\textit{After}}_t + {\textit{controls}} + {e}_{\textit{it}}$$ Variables (1) OLS no controls (2) OLS with controls (3) BIG 5 (4) IV AA1995 (5) IV AA9899 (6) INTRADAY (7) RATED (8) UNRATED (9) RATED 2-day (10) UNRATED 2-day AAClient*After –0.665** –0.630** –0.720** –0.519* –0.768** –0.555*** 0.067 –1.194*** 0.315 –0.781 (0.291) (0.315) (0.312) (0.285) (0.340) (0.215) $$\underbrace {\matrix{{\left( {0.405} \right)} & {\left( {0.453} \right)} \cr} }_{ - {{1.261}^{ * * }}}$$ $$\underbrace {\matrix{{\left( {0.574} \right)} & {\left( {0.781} \right)} \cr } }_{ - 1.096}$$ (0.608) (0.965) AAClient 0.271 0.220 0.259 0.286 0.331 0.455*** –0.110 0.500 –0.241 0.095 (0.190) (0.212) (0.204) (0.199) (0.236) (0.135) (0.249) (0.312) (0.390) (0.585) After 0.332** (0.139) log(Asset) 0.075 0.083 –0.003 0.023 0.116** –0.169 0.259 0.007 0.672** (0.083) (0.080) (0.084) (0.084) (0.052) (0.112) (0.165) (0.152) (0.313) Book-to-market 0.085 0.106 –0.010 0.084 0.051 0.117 0.042 0.083 –0.208 (0.096) (0.095) (0.106) (0.102) (0.076) (0.135) (0.134) (0.198) (0.235) Market leverage –0.104 –0.124 0.034 –0.051 –0.117* –0.182 –0.156 –0.310 –0.115 (0.105) (0.103) (0.122) (0.109) (0.068) (0.189) (0.147) (0.202) (0.268) Profitability 0.066 0.027 0.161 0.217** –0.035 –0.142 0.027 –0.228 –0.013 (0.095) (0.095) (0.140) (0.096) (0.058) (0.285) (0.106) (0.274) (0.184) CAPM beta 0.320*** 0.313*** 0.292* 0.353*** 0.343*** 0.333 0.242* 0.013 0.556** (0.115) (0.117) (0.177) (0.135) (0.077) (0.221) (0.140) (0.300) (0.222) Constant 0.685*** (0.085) Number of obs. 2,776 2,554 2,452 1,725 2,199 1,902 970 1,584 968 1,553 R-squared 0.003 0.027 0.028 0.030 0.031 0.121 0.058 0.029 0.061 0.044 All regressions, except for (1), include nine SIC industry fixed effects and their interaction with After. Hence the coefficient of Constant and After is not reported for these regressions. All reported coefficients of control variables are standardized. The dependent variable is from CRSP (daily) or QuantQuote (intraday) and expressed in percentage points. The entry between Columns 7 and 8 is the estimate for the difference in the two subsamples. Heteroscedasticity-robust standard errors are in parentheses. Standard errors clustered at the auditor level and block-bootstrapped standard errors were smaller than heteroscedasticity-robust errors; hence, heteroscedasticity-robust errors are reported throughout. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. After=0 refers to May 15, 2001, and After=1 refers to November 6, 2001. The dependent variable is the daily stock returns on May 15, 2001, and November 6, 2001, from CRSP. All balance sheet variables used for calculating control variables are described in the notes of Table 5. In Column 3, BIG 5 refers to the subsample of Big Five auditing firms’ clients. Column 4 is the instrumental variable regression with auditor choice in 1995 used as the instrument. The Hausman $$\chi^2$$ test statistic, equal to 5.20 ($$\text{p}=0.74$$), comes from the bootstrapped Hausman test, as in Cameron, and Trivedi (2010, p. 443), that does not require one of the estimators to be efficient. Column 6 is the instrumental variable regressions with the auditor choice of 1998 and 1999 used as the instruments. Table 6 The ENRON scandal’s effect on stock price sensitivity to monetary policy Arthur Andersen Clients versus Other Firms: May 15, 2001 versus November 6, 2001 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 + {\textit{AAClient}}_i + \beta_2 * {\textit{After}}_t + \tilde{\delta} * {\textit{AAClient}}_i + {\textit{After}}_t + {\textit{controls}} + {e}_{\textit{it}}$$ Variables (1) OLS no controls (2) OLS with controls (3) BIG 5 (4) IV AA1995 (5) IV AA9899 (6) INTRADAY (7) RATED (8) UNRATED (9) RATED 2-day (10) UNRATED 2-day AAClient*After –0.665** –0.630** –0.720** –0.519* –0.768** –0.555*** 0.067 –1.194*** 0.315 –0.781 (0.291) (0.315) (0.312) (0.285) (0.340) (0.215) $$\underbrace {\matrix{{\left( {0.405} \right)} & {\left( {0.453} \right)} \cr} }_{ - {{1.261}^{ * * }}}$$ $$\underbrace {\matrix{{\left( {0.574} \right)} & {\left( {0.781} \right)} \cr } }_{ - 1.096}$$ (0.608) (0.965) AAClient 0.271 0.220 0.259 0.286 0.331 0.455*** –0.110 0.500 –0.241 0.095 (0.190) (0.212) (0.204) (0.199) (0.236) (0.135) (0.249) (0.312) (0.390) (0.585) After 0.332** (0.139) log(Asset) 0.075 0.083 –0.003 0.023 0.116** –0.169 0.259 0.007 0.672** (0.083) (0.080) (0.084) (0.084) (0.052) (0.112) (0.165) (0.152) (0.313) Book-to-market 0.085 0.106 –0.010 0.084 0.051 0.117 0.042 0.083 –0.208 (0.096) (0.095) (0.106) (0.102) (0.076) (0.135) (0.134) (0.198) (0.235) Market leverage –0.104 –0.124 0.034 –0.051 –0.117* –0.182 –0.156 –0.310 –0.115 (0.105) (0.103) (0.122) (0.109) (0.068) (0.189) (0.147) (0.202) (0.268) Profitability 0.066 0.027 0.161 0.217** –0.035 –0.142 0.027 –0.228 –0.013 (0.095) (0.095) (0.140) (0.096) (0.058) (0.285) (0.106) (0.274) (0.184) CAPM beta 0.320*** 0.313*** 0.292* 0.353*** 0.343*** 0.333 0.242* 0.013 0.556** (0.115) (0.117) (0.177) (0.135) (0.077) (0.221) (0.140) (0.300) (0.222) Constant 0.685*** (0.085) Number of obs. 2,776 2,554 2,452 1,725 2,199 1,902 970 1,584 968 1,553 R-squared 0.003 0.027 0.028 0.030 0.031 0.121 0.058 0.029 0.061 0.044 Arthur Andersen Clients versus Other Firms: May 15, 2001 versus November 6, 2001 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 + {\textit{AAClient}}_i + \beta_2 * {\textit{After}}_t + \tilde{\delta} * {\textit{AAClient}}_i + {\textit{After}}_t + {\textit{controls}} + {e}_{\textit{it}}$$ Variables (1) OLS no controls (2) OLS with controls (3) BIG 5 (4) IV AA1995 (5) IV AA9899 (6) INTRADAY (7) RATED (8) UNRATED (9) RATED 2-day (10) UNRATED 2-day AAClient*After –0.665** –0.630** –0.720** –0.519* –0.768** –0.555*** 0.067 –1.194*** 0.315 –0.781 (0.291) (0.315) (0.312) (0.285) (0.340) (0.215) $$\underbrace {\matrix{{\left( {0.405} \right)} & {\left( {0.453} \right)} \cr} }_{ - {{1.261}^{ * * }}}$$ $$\underbrace {\matrix{{\left( {0.574} \right)} & {\left( {0.781} \right)} \cr } }_{ - 1.096}$$ (0.608) (0.965) AAClient 0.271 0.220 0.259 0.286 0.331 0.455*** –0.110 0.500 –0.241 0.095 (0.190) (0.212) (0.204) (0.199) (0.236) (0.135) (0.249) (0.312) (0.390) (0.585) After 0.332** (0.139) log(Asset) 0.075 0.083 –0.003 0.023 0.116** –0.169 0.259 0.007 0.672** (0.083) (0.080) (0.084) (0.084) (0.052) (0.112) (0.165) (0.152) (0.313) Book-to-market 0.085 0.106 –0.010 0.084 0.051 0.117 0.042 0.083 –0.208 (0.096) (0.095) (0.106) (0.102) (0.076) (0.135) (0.134) (0.198) (0.235) Market leverage –0.104 –0.124 0.034 –0.051 –0.117* –0.182 –0.156 –0.310 –0.115 (0.105) (0.103) (0.122) (0.109) (0.068) (0.189) (0.147) (0.202) (0.268) Profitability 0.066 0.027 0.161 0.217** –0.035 –0.142 0.027 –0.228 –0.013 (0.095) (0.095) (0.140) (0.096) (0.058) (0.285) (0.106) (0.274) (0.184) CAPM beta 0.320*** 0.313*** 0.292* 0.353*** 0.343*** 0.333 0.242* 0.013 0.556** (0.115) (0.117) (0.177) (0.135) (0.077) (0.221) (0.140) (0.300) (0.222) Constant 0.685*** (0.085) Number of obs. 2,776 2,554 2,452 1,725 2,199 1,902 970 1,584 968 1,553 R-squared 0.003 0.027 0.028 0.030 0.031 0.121 0.058 0.029 0.061 0.044 All regressions, except for (1), include nine SIC industry fixed effects and their interaction with After. Hence the coefficient of Constant and After is not reported for these regressions. All reported coefficients of control variables are standardized. The dependent variable is from CRSP (daily) or QuantQuote (intraday) and expressed in percentage points. The entry between Columns 7 and 8 is the estimate for the difference in the two subsamples. Heteroscedasticity-robust standard errors are in parentheses. Standard errors clustered at the auditor level and block-bootstrapped standard errors were smaller than heteroscedasticity-robust errors; hence, heteroscedasticity-robust errors are reported throughout. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. After=0 refers to May 15, 2001, and After=1 refers to November 6, 2001. The dependent variable is the daily stock returns on May 15, 2001, and November 6, 2001, from CRSP. All balance sheet variables used for calculating control variables are described in the notes of Table 5. In Column 3, BIG 5 refers to the subsample of Big Five auditing firms’ clients. Column 4 is the instrumental variable regression with auditor choice in 1995 used as the instrument. The Hausman $$\chi^2$$ test statistic, equal to 5.20 ($$\text{p}=0.74$$), comes from the bootstrapped Hausman test, as in Cameron, and Trivedi (2010, p. 443), that does not require one of the estimators to be efficient. Column 6 is the instrumental variable regressions with the auditor choice of 1998 and 1999 used as the instruments. The necessary and sufficient identification assumption for Columns 1 and 2 is that the auditor choice is independent of the characteristics that might affect how a firm’s stock price sensitivity to monetary policy shocks changes with increasing information frictions. However, the perceived auditing quality between clients of Big Five auditing firms and the remaining auditing firms might have been different in 2000. Therefore, Column 3 repeats the regression by focusing on the subsample of Big Five auditing firms’ clients. Because the firms that prepare clean balance sheets, and hence have nonmissing control variables, tend to be clients of the Big Five accounting firms, the number of observations in Column 3 differs little from the number of observations in Column 2, and the coefficient estimates are very close to each other, as expected. In order to address any remaining concerns about endogeneity, I note that firms usually establish long-term relationships with their auditors. Correspondingly, I use the auditor choice in fiscal year 1995 as an instrument in an instrumental variable (IV) framework. By using a long lag of five years (1995 vs. 2000), the IV approach tries to be conservative to stack the odds against the main hypothesis. While popular, using lagged independent variables as instruments can still cause problems if the source of endogeneity, such as omitted variables, is also persistent (Roberts and Whited (2013)). Using a long lag (1995 vs. 2000) alleviates this problem for this particular case, because a firm’s auditor choice is much more persistent than many other firm characteristics that could potentially affect the reaction of stock prices to monetary policy shocks as the Enron scandal unfolds.18 Column 4 shows that the resultant coefficient for AAClient*After has the same sign and a similar magnitude as before. The Hausman $$\chi ^{2}$$ test statistic, equal to $$5.20$$$$(p=0.74)$$, suggests that endogeneity is not a big concern. Finally, I find that after controlling for the auditor choice in 1998 and 1999, the auditor choice in earlier years does not have a significant explanatory power for the auditor choice in 2000. Therefore, for completeness, Column 5 reports the instrumental variable estimates using the auditor choice in 1998 and 1999 as instruments, and these estimates are similar to the previous estimates. All of these instruments pass the standard tests for instrument weakness and overidentification. The Achilles heel of the difference-in-differences approach is the nonparallel time trend across firms. In order to address this immediate concern, I separate stocks into portfolios by their auditing firms and run a regression of different portfolio returns on a linear trend. I find that there is no significant difference in the coefficients, which suggests that a nonparallel monotonic time trend is not a primary concern (see the Online Appendix). Nevertheless, a linear time trend might not capture a confounding factor, such as news related to Enron. Therefore, I use intraday returns between 2:00 p.m. (15 minutes before the FOMC announcement) and 4:00 p.m. (the market close) to ensure that most of the stock return movement is attributable to the monetary policy announcement.19 Column 6 in Table 6 shows that, for the intraday returns, the coefficient of AAClient*After is of a magnitude similar to the coefficients in Columns 1 and 2 (differing from each other only by about half a standard deviation) and is still statistically significant. Columns 7 and 8 separately present the results for the rated and unrated firms. The first lines in these columns show that, for these two groups of firms, the estimated coefficients of interest differ from each other in an economically significant way, and the second line between these columns shows that this difference ($$-1.26$$) is statistically significant as well. In particular, almost all of the effect seems to stem from unrated firms which suggests that financial statements are a more important source of information for unrated firms, consistent with the common notion that unrated firms are more opaque.20 Finally, Columns 9 and 10 show that the economic magnitude of the difference between rated and unrated firms survive when we look at returns over a 2-day interval but it becomes statistically insignificant. This result is consistent with the preference of earlier papers to study daily or intraday returns for better identification of the effect of monetary policy, a preference followed in this paper as well. 2.2.1 Panel data and additional robustness checks The difference-in-differences approach above uses only two policy announcement dates in 2001 because it is straightforward to argue that investors were much more certain about the accounting scandal on November 6 compared to May 15. If there were a direct measure of investors’ beliefs about the scandal over the course of 2001, we could also use the other policy announcement dates in 2001 in a panel setting. Despite the absence of such a direct measure, one can still attempt a panel data analysis if one is content with a proxy for investors’ beliefs about the scandal. It is realistic to assume that Enron’s fate in 2001 was tied to the outcome of the accounting scandal involving Arthur Andersen. Therefore, one can use Enron’s proximity to default as a proxy for investors’ beliefs. I use Moody’s expected default frequency (EDF) of Enron as this proxy. The EDF is a measure of the probability that a firm will default over a specified period of time (typically one year) where default is defined as failure to make scheduled principal or interest payments.21 This approach also addresses potential concerns due to minor differences in monetary policy surprises on the two dates used by the difference-in-differences analysis and validates the previous results. The event dates consist of the scheduled FOMC announcement dates in 2001, which include January 31, March 20, May 15, June 27, August 21, October 2, November 6, and December 11. The sample stops after 2001 because 2002 was riddled with accounting scandals involving other auditing firms, starting in January with Homestore.com whose auditing firm was Pricewaterhouse-Coopers, and in February with Qwest, whose auditing firm was KPMG. The policy surprises in 2001 are sizeable, with average absolute surprise of 4.75 basis points compared to 3.35 bp for the rest of the 1994–2008 sample, which makes this period particularly useful for identification. The econometric model is specified as \begin{eqnarray} return_{it} &=&\beta _{0}+\beta _{1}\text{PolicySurprise}_{t}\ast \text{AAClient}_{i}+\delta _{1}\text{EDF}_{t}\ast \text{AAClient}_{i} \nonumber \\ &&+\,\delta _{2}\text{EDF}_{t}\ast \text{PolicySurprise}_{t}\ast \text{AAClient}_{i} \\ &&+\,\text{time fixed effects}+\text{firm fixed effects} \nonumber \\ &&+\,\text{other controls}+e_{it}\text{,} \nonumber \end{eqnarray} (7) where PolicySurprise is the monetary policy surprise on the scheduled FOMC announcement dates in 2001, as defined in Section 1. The standalone AAClient dummy is absorbed by the firm fixed effect, and the standalone EDF, PolicySurprise, and their interaction, EDF*(PolicySurprise), are absorbed by the date fixed effects. The EDF is calculated by first taking the logarithm of Moody’s daily expected default frequency of Enron in order to control for nonlinearities and then taking its equally weighted 10-day moving average (up to, but not including, the FOMC dates) in order to reduce the mismeasurement due to high volatility of the daily EDF measure.22 The parameter of interest is $$\delta _{2}$$; that is, how Enron’s proximity to default affects the relative stock price reaction of Arthur Andersen clients to monetary policy surprises. Our previous results suggest that $$\delta _{2}<0$$. The results in Table 7 are consistent with the results from the difference-in-differences analysis and the conjecture that ties Enron’s fate to the accounting scandal. In particular, as Enron’s proximity to default increases, the stock prices of Arthur Andersen clients react less to monetary policy surprises in comparison to the stock prices of other auditing firms’ clients. The results in Table 7 can be compared to those in Table 6 if the estimates in Table 7 are used in order to calculate the relative change in the monetary policy sensitivity of Arthur Andersen clients’ stock prices from May 15 to November 6, 2001. During this period, Enron’s EDF measure changed by $$2.79$$. Therefore, Column 1 in Table 7 implies that a 10 basis point surprise decline in the federal funds target rate generates a reaction that is about $$50$$ bp ($$1.71\ast \left( \Delta EDF\right) \ast 10$$) less for the stock prices of Arthur Andersen clients compared to the clients of other auditing firms, which is in the ballpark of the numbers reported in Table 6. Moreover, Columns 2 and 3 in Table 7 imply that the difference between rated and unrated firms is $$-6.29\ast \left( \Delta EDF\right) \ast 10\approx -175$$ bp ($$-1.75$$ percentage points), which is consistent with the $$-1.26$$$$(-1.194-0.067)$$ percentage point estimate reported in Table 6. Table 7 The ENRON scandal’s effect on stock price sensitivity to monetary policy Arthur Andersen clients versus other firms: The eight scheduled FOMC announcement dates in 2001 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \delta_1 * {\textit{EDF}}_t * {\textit{AAClient}}_i + \delta_2 * {\textit{EDF}}_t * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \text{controls}_i + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Variables ALL RATED UNRATED BIG SMALL LOW R&D HIGH R&D ACCR $$<=0$$ ACCR $$>0$$ ACC2 $$<=0$$ ACC2 $$>0$$ EDF*PolicySurprise –1.71 1.99 –4.30*** 0.53 –3.87** 1.23 –8.43*** –0.71 –7.09** –0.66 –8.12*** *AAClient (1.15) $$\underbrace {\matrix{{\left( {1.60} \right)} & {\left( {1.55} \right)} \cr } }_{ - {{6.29}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.50} \right)} & {\left( {1.72} \right)} \cr } }_{ - {{4.40}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.67} \right)} & {\left( {2.72} \right)} \cr} }_{ - {{9.96}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.44} \right)} & {\left( {2.80} \right)} \cr } }_{ - {{6.38}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.58} \right)} & {\left( {2.93} \right)} \cr } }_{ - {{7.46}^{ * * }}}$$ Difference between groups (2.23) (2.28) (3.19) (3.14) (3.33) PolicySurprise –1.25 –0.51 –1.70 –2.48 –0.09 –0.02 –0.37 –1.18 –1.84 –1.87 –0.75 *AAClient (1.50) (1.88) (2.14) (1.68) (2.43) (2.31) (4.15) (1.83) (4.04) (2.06) (4.50) EDF*AAClient –4.59 –13.60* 1.72 –14.15* 4.54 –21.08* 15.39 –10.78 22.39 –13.04 35.04* (6.94) (8.10) (10.15) (7.90) (11.21) (11.74) (19.09) (8.00) (17.90) (9.36) (18.89) Number of observations 11,382 4,037 7,345 5,371 6,011 2,880 3,121 6,913 2,112 5,681 1,923 R-squared 0.05 0.08 0.05 0.09 0.04 0.04 0.10 0.06 0.05 0.06 0.05 Number of stocks 1,754 542 1,212 720 1,034 430 540 1,047 328 878 303 Arthur Andersen clients versus other firms: The eight scheduled FOMC announcement dates in 2001 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \delta_1 * {\textit{EDF}}_t * {\textit{AAClient}}_i + \delta_2 * {\textit{EDF}}_t * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \text{controls}_i + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Variables ALL RATED UNRATED BIG SMALL LOW R&D HIGH R&D ACCR $$<=0$$ ACCR $$>0$$ ACC2 $$<=0$$ ACC2 $$>0$$ EDF*PolicySurprise –1.71 1.99 –4.30*** 0.53 –3.87** 1.23 –8.43*** –0.71 –7.09** –0.66 –8.12*** *AAClient (1.15) $$\underbrace {\matrix{{\left( {1.60} \right)} & {\left( {1.55} \right)} \cr } }_{ - {{6.29}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.50} \right)} & {\left( {1.72} \right)} \cr } }_{ - {{4.40}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.67} \right)} & {\left( {2.72} \right)} \cr} }_{ - {{9.96}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.44} \right)} & {\left( {2.80} \right)} \cr } }_{ - {{6.38}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.58} \right)} & {\left( {2.93} \right)} \cr } }_{ - {{7.46}^{ * * }}}$$ Difference between groups (2.23) (2.28) (3.19) (3.14) (3.33) PolicySurprise –1.25 –0.51 –1.70 –2.48 –0.09 –0.02 –0.37 –1.18 –1.84 –1.87 –0.75 *AAClient (1.50) (1.88) (2.14) (1.68) (2.43) (2.31) (4.15) (1.83) (4.04) (2.06) (4.50) EDF*AAClient –4.59 –13.60* 1.72 –14.15* 4.54 –21.08* 15.39 –10.78 22.39 –13.04 35.04* (6.94) (8.10) (10.15) (7.90) (11.21) (11.74) (19.09) (8.00) (17.90) (9.36) (18.89) Number of observations 11,382 4,037 7,345 5,371 6,011 2,880 3,121 6,913 2,112 5,681 1,923 R-squared 0.05 0.08 0.05 0.09 0.04 0.04 0.10 0.06 0.05 0.06 0.05 Number of stocks 1,754 542 1,212 720 1,034 430 540 1,047 328 878 303 All regressions include firm and time fixed effects. Errors in parentheses are clustered at the firm level. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. Both the dependent variable (returns) and the surprise component of the federal funds target rate change are expressed in basis points. The numbers between Columns 2 and 3, 4 and 5, 6 and 7, 8 and 9, and 10 and 11 give the estimates for the difference between the two corresponding subsamples. PolicySurprise is scaled so that positive surprises indicate expansionary shocks. To save space, the results for other control variables are not reported. The calculation of firm-level controls is discussed in Table 5. EDF is the 10-day moving average of the log of Moody’s expected default frequency for Enron. The AAClient dummy is absorbed by the firm fixed effect, and EDF, PolicySurprise, and EDF*PolicySurprise are absorbed by the date fixed effects. ACCR is the ratio of accruals to total assets as in Sufi (2007), and ACC2 is the same object where accruals also include Equity in Earnings – Unconsolidated Subsidiaries (Compustat item ESUB). BIG versus SMALL and HIGH R&D versus LOW R&D are determined using the median of AT and XRD/AT from Compustat, respectively. The announcement dates are January 31st, March 20th, May 15th, June 27th, August 21st, October 2nd, November 6th, and December 11th. Table 7 The ENRON scandal’s effect on stock price sensitivity to monetary policy Arthur Andersen clients versus other firms: The eight scheduled FOMC announcement dates in 2001 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \delta_1 * {\textit{EDF}}_t * {\textit{AAClient}}_i + \delta_2 * {\textit{EDF}}_t * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \text{controls}_i + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Variables ALL RATED UNRATED BIG SMALL LOW R&D HIGH R&D ACCR $$<=0$$ ACCR $$>0$$ ACC2 $$<=0$$ ACC2 $$>0$$ EDF*PolicySurprise –1.71 1.99 –4.30*** 0.53 –3.87** 1.23 –8.43*** –0.71 –7.09** –0.66 –8.12*** *AAClient (1.15) $$\underbrace {\matrix{{\left( {1.60} \right)} & {\left( {1.55} \right)} \cr } }_{ - {{6.29}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.50} \right)} & {\left( {1.72} \right)} \cr } }_{ - {{4.40}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.67} \right)} & {\left( {2.72} \right)} \cr} }_{ - {{9.96}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.44} \right)} & {\left( {2.80} \right)} \cr } }_{ - {{6.38}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.58} \right)} & {\left( {2.93} \right)} \cr } }_{ - {{7.46}^{ * * }}}$$ Difference between groups (2.23) (2.28) (3.19) (3.14) (3.33) PolicySurprise –1.25 –0.51 –1.70 –2.48 –0.09 –0.02 –0.37 –1.18 –1.84 –1.87 –0.75 *AAClient (1.50) (1.88) (2.14) (1.68) (2.43) (2.31) (4.15) (1.83) (4.04) (2.06) (4.50) EDF*AAClient –4.59 –13.60* 1.72 –14.15* 4.54 –21.08* 15.39 –10.78 22.39 –13.04 35.04* (6.94) (8.10) (10.15) (7.90) (11.21) (11.74) (19.09) (8.00) (17.90) (9.36) (18.89) Number of observations 11,382 4,037 7,345 5,371 6,011 2,880 3,121 6,913 2,112 5,681 1,923 R-squared 0.05 0.08 0.05 0.09 0.04 0.04 0.10 0.06 0.05 0.06 0.05 Number of stocks 1,754 542 1,212 720 1,034 430 540 1,047 328 878 303 Arthur Andersen clients versus other firms: The eight scheduled FOMC announcement dates in 2001 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \delta_1 * {\textit{EDF}}_t * {\textit{AAClient}}_i + \delta_2 * {\textit{EDF}}_t * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \text{controls}_i + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Variables ALL RATED UNRATED BIG SMALL LOW R&D HIGH R&D ACCR $$<=0$$ ACCR $$>0$$ ACC2 $$<=0$$ ACC2 $$>0$$ EDF*PolicySurprise –1.71 1.99 –4.30*** 0.53 –3.87** 1.23 –8.43*** –0.71 –7.09** –0.66 –8.12*** *AAClient (1.15) $$\underbrace {\matrix{{\left( {1.60} \right)} & {\left( {1.55} \right)} \cr } }_{ - {{6.29}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.50} \right)} & {\left( {1.72} \right)} \cr } }_{ - {{4.40}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.67} \right)} & {\left( {2.72} \right)} \cr} }_{ - {{9.96}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.44} \right)} & {\left( {2.80} \right)} \cr } }_{ - {{6.38}^{ * * * }}}$$ $$\underbrace {\matrix{{\left( {1.58} \right)} & {\left( {2.93} \right)} \cr } }_{ - {{7.46}^{ * * }}}$$ Difference between groups (2.23) (2.28) (3.19) (3.14) (3.33) PolicySurprise –1.25 –0.51 –1.70 –2.48 –0.09 –0.02 –0.37 –1.18 –1.84 –1.87 –0.75 *AAClient (1.50) (1.88) (2.14) (1.68) (2.43) (2.31) (4.15) (1.83) (4.04) (2.06) (4.50) EDF*AAClient –4.59 –13.60* 1.72 –14.15* 4.54 –21.08* 15.39 –10.78 22.39 –13.04 35.04* (6.94) (8.10) (10.15) (7.90) (11.21) (11.74) (19.09) (8.00) (17.90) (9.36) (18.89) Number of observations 11,382 4,037 7,345 5,371 6,011 2,880 3,121 6,913 2,112 5,681 1,923 R-squared 0.05 0.08 0.05 0.09 0.04 0.04 0.10 0.06 0.05 0.06 0.05 Number of stocks 1,754 542 1,212 720 1,034 430 540 1,047 328 878 303 All regressions include firm and time fixed effects. Errors in parentheses are clustered at the firm level. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. Both the dependent variable (returns) and the surprise component of the federal funds target rate change are expressed in basis points. The numbers between Columns 2 and 3, 4 and 5, 6 and 7, 8 and 9, and 10 and 11 give the estimates for the difference between the two corresponding subsamples. PolicySurprise is scaled so that positive surprises indicate expansionary shocks. To save space, the results for other control variables are not reported. The calculation of firm-level controls is discussed in Table 5. EDF is the 10-day moving average of the log of Moody’s expected default frequency for Enron. The AAClient dummy is absorbed by the firm fixed effect, and EDF, PolicySurprise, and EDF*PolicySurprise are absorbed by the date fixed effects. ACCR is the ratio of accruals to total assets as in Sufi (2007), and ACC2 is the same object where accruals also include Equity in Earnings – Unconsolidated Subsidiaries (Compustat item ESUB). BIG versus SMALL and HIGH R&D versus LOW R&D are determined using the median of AT and XRD/AT from Compustat, respectively. The announcement dates are January 31st, March 20th, May 15th, June 27th, August 21st, October 2nd, November 6th, and December 11th. Regarding this result, an obvious concern is that larger firms are more likely to be rated, and firm size, rather than ratings availability, is the main driver of these results. Indeed, Columns 4 and 5 show that if the sample is divided into large and small firms based on the median firm size, a pattern similar to that for rated and unrated firms emerges. In order to address this concern, the sample is divided into four groups based on size and rating availability, and the coefficients on EDF$$\ast $$PolicySurprise$$\ast $$AAClient for each group are compared in Table 8. As expected, the majority of the sample is concentrated in the Big&Rated and Small&Unrated groups. The table’s message is clear: being an Arthur Andersen client has a more significant effect among unrated firms compared to rated firms and this result is economically large and statistically significant regardless of whether we look at the group of big firms, where the difference is $$-$$6.37, or at the group of small firms, where the difference is $$-$$8.96. However, being an Arthur Andersen client has no significant effect on small firms compared to large firms once the rating status of the firms is taken into account; if anything the effect actually goes in the opposite direction for size as it can be seen in the last two lines of the table. These results suggest that the effect of ratings availability dominates the effect of firm size. Table 8 Coefficient of EDF*PolicySurprise*AAClient when firms are double sorted \begin{array}{rcl} \textit{return}_{\textit{it}}&=&\beta_0 + \beta_1 * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \delta_1 * {\textit{EDF}}_t * {\textit{AAClient}}_i\\ & &+ \delta_2 * {\textit{EDF}}_t * {\textit{PolicySurprise}}_t * \text{AAClient}_i + \text{controls} + {e}_{\textit{it}}\end{array} RATED UNRATED UNRATED-RATED difference BIG 1.793 –4.571 –6.365* (1.681) (2.965) (3.401) # firms 498 222 SMALL 4.671 –4.288** –8.960** (4.055) (1.792) (4.333) # firms 44 990 SMALL-BIG difference 2.877 0.282 (4.292) (3.456) \begin{array}{rcl} \textit{return}_{\textit{it}}&=&\beta_0 + \beta_1 * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \delta_1 * {\textit{EDF}}_t * {\textit{AAClient}}_i\\ & &+ \delta_2 * {\textit{EDF}}_t * {\textit{PolicySurprise}}_t * \text{AAClient}_i + \text{controls} + {e}_{\textit{it}}\end{array} RATED UNRATED UNRATED-RATED difference BIG 1.793 –4.571 –6.365* (1.681) (2.965) (3.401) # firms 498 222 SMALL 4.671 –4.288** –8.960** (4.055) (1.792) (4.333) # firms 44 990 SMALL-BIG difference 2.877 0.282 (4.292) (3.456) Each cell goes from top to bottom: the coefficient, standard error, and number of firms. BIG versus SMALL are calculated using the median of total assets (AT) from Compustat. RATED and UNRATED are defined as in Table 1. Table 8 Coefficient of EDF*PolicySurprise*AAClient when firms are double sorted \begin{array}{rcl} \textit{return}_{\textit{it}}&=&\beta_0 + \beta_1 * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \delta_1 * {\textit{EDF}}_t * {\textit{AAClient}}_i\\ & &+ \delta_2 * {\textit{EDF}}_t * {\textit{PolicySurprise}}_t * \text{AAClient}_i + \text{controls} + {e}_{\textit{it}}\end{array} RATED UNRATED UNRATED-RATED difference BIG 1.793 –4.571 –6.365* (1.681) (2.965) (3.401) # firms 498 222 SMALL 4.671 –4.288** –8.960** (4.055) (1.792) (4.333) # firms 44 990 SMALL-BIG difference 2.877 0.282 (4.292) (3.456) \begin{array}{rcl} \textit{return}_{\textit{it}}&=&\beta_0 + \beta_1 * {\textit{PolicySurprise}}_t * {\textit{AAClient}}_i + \delta_1 * {\textit{EDF}}_t * {\textit{AAClient}}_i\\ & &+ \delta_2 * {\textit{EDF}}_t * {\textit{PolicySurprise}}_t * \text{AAClient}_i + \text{controls} + {e}_{\textit{it}}\end{array} RATED UNRATED UNRATED-RATED difference BIG 1.793 –4.571 –6.365* (1.681) (2.965) (3.401) # firms 498 222 SMALL 4.671 –4.288** –8.960** (4.055) (1.792) (4.333) # firms 44 990 SMALL-BIG difference 2.877 0.282 (4.292) (3.456) Each cell goes from top to bottom: the coefficient, standard error, and number of firms. BIG versus SMALL are calculated using the median of total assets (AT) from Compustat. RATED and UNRATED are defined as in Table 1. The rest of Table 7 uses alternative measures of firm opaqueness. Columns 6 and 7 provide a comparison of firms with low and high R&D spending (relative to total assets) and find a very similar pattern, but one that is actually stronger in terms of its magnitude when compared to the difference between rated and unrated firms ($$-9.66$$ versus $$-6.29$$). Remaining columns repeat the same analysis for firms with positive accruals, as these firms are more likely to engage in earnings management and hence need better auditing practices. Two different accruals measures are used. The first measure comes from Sloan (1996), which is also used in Sufi (2007). The second measure modifies this first measure by adding accruals (unremitted earnings) from unconsolidated subsidiaries (Compustat item ESUB) because off-balance-sheet assets played a very significant role in the Enron scandal, which, in turn, might have drawn investors’ attention to these assets. Besides addressing the earnings management issue that is directly relevant to the reliability of financial statements, accruals have the additional advantage of providing extra credibility to the results for the main hypothesis because they are actually negatively correlated with R&D spending in my sample, with Pearson and Spearman correlations of $$-$$0.2 and $$-$$0.1, respectively. Columns 8 and 9 provide the results for the Sloan accruals measure, and Columns 10 and 11 provide the results for the modified accruals measure. Both sets of results are consistent with the results that are obtained using ratings availability and R&D spending. In particular, the differential effect between Arthur Andersen clients and other firms seems to stem from firms that are more likely to engage in earnings management because these firms’ information frictions take a bigger hit if their auditor’s reputation is tarnished. As seen in the second line between Columns 8 and 9 and between Columns 10 and 11, this result is both economically and statistically significant like the result for the R&D measure in Columns 6 and 7, despite the slightly negative cross-sectional correlation between the accrual measure and the R&D spending measure. The Online Appendix provides a fully saturated model with over 100 additional interaction terms of industry dummies and firm-level controls with EDF and EDF*PolicySurprise and shows that the results remain similar. Other robustness checks include the following: the replacement of Enron’s EDF with its idiosyncratic stock price movements to remove the potential impact of other macroeconomic events that may have affected Enron over the course of 2001, the use of Enron’s stock returns on FOMC dates to control for any Enron-related news on these dates, and a placebo experiment using the FOMC dates in 2000. None of these robustness checks alters the main results of this section. The Online Appendix also shows that the effect of the reliability of financial statements is not limited to the Enron scandal, using the financial restatements database of the Government Accountability Office. In particular, the stock price of a restating firm becomes less sensitive to monetary policy surprises following the announcement of financial restatement, by about 40 bp for a 10-bp surprise change. 3. Explaining the Importance of Information Frictions My results so far conclude that firms subject to greater information frictions have stock prices that are less responsive to monetary policy. While there may be multiple channels that can explain this pattern, one particular channel seems to stand out both in the simple model in the appendix and in the data. In particular, information frictions may lead to lower leverage, which in turn reduces the responsiveness of stock prices to monetary policy. This section will review the evidence regarding the differences in leverage between financially constrained and unconstrained firms and how well these differences fit together with the differences in the monetary policy sensitivity of different stocks. I will also discuss other potential mechanisms at the end of this section. I start with regressions similar to those in Section 1 in order to establish how financial frictions can lead to lower monetary policy sensitivity of stock prices through lower leverage. The evidence from the Enron scandal (Figure 2) has shown us that an event that changes a firm’s access to financing can take years to show up in the firm’s capital structure although the effect on stock prices is immediate. In the words of Graham, Hughson, and Zender (1999), “if the capital structure change is delayed due to transaction costs, there is no guarantee that the timing of the stock price reaction and the capital structure change will coincide.” The advantage of financial constraint and opaqueness proxies is that they are more forward looking because, for example, a change in ratings availability has an immediate effect on investors’ expectations regarding a firm’s future access to credit. Moreover, these proxies are less rigid because variables that are rapidly responding to a firm’s financial condition have an immediate effect on them, such as the effect of income or dividends on the Whited-Wu index. Therefore, current book leverage may be more prone to measurement error as a measure of future access to credit than financial friction proxies that are more forward-looking or fast-changing. Consistent with this argument, Column 1 of Table 9 shows that when we directly look at the effect of book leverage on the monetary policy sensitivity of stock prices, the effect is relatively small: a one standard deviation (0.2) lower book leverage leads to about 0.2 percentage reduction in the responsiveness of stock prices to monetary policy. However, when we use the various financial constraint proxies from Section 1 that seem to be strongly related to monetary policy sensitivity of stock prices as instruments for book leverage, we find that the effect is much stronger. This is true both when each proxy is used alone (Columns 2–5) or when all proxies are used together as instruments as in the last column (Column 6). Overall, a 1 standard deviation (0.2) lower book leverage due to financial frictions leads to about 1.3 (6.44*0.2) percentage reduction in the responsiveness of stock prices to monetary policy.23 Table 9 Leverage and stock price sensitivity to monetary policy $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 * {\textit{PolicySurprise}}_t + \beta_2 * {\textit{Leverage}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Leverage}}_{\textit{it}-1} + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) (6) FE IV IV IV IV IV Variables UNRATED ACC>0 WW HP ALL PolicySurprise 1.99*** 0.96*** –0.44 –1.02 1.66* 1.58*** (0.38) (0.35) (2.15) (1.29) (0.88) (0.48) PolicySurprise*Leverage 1.00* 8.29*** 13.97 17.60*** 5.19 6.44*** (0.60) (1.37) (8.91) (5.34) (3.70) (1.89) Leverage 0.08 –0.16** 1.45** –0.68*** –0.37* –0.33*** (0.07) (0.08) (0.62) (0.26) (0.20) (0.10) Constant –0.18 0.27*** –0.10 0.42*** 0.31*** 0.34*** (1.44) (0.02) (0.15) (0.06) (0.05) (0.03) Number of obs. 444,647 444,718 315,896 256,259 444,647 246,572 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 * {\textit{PolicySurprise}}_t + \beta_2 * {\textit{Leverage}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Leverage}}_{\textit{it}-1} + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) (6) FE IV IV IV IV IV Variables UNRATED ACC>0 WW HP ALL PolicySurprise 1.99*** 0.96*** –0.44 –1.02 1.66* 1.58*** (0.38) (0.35) (2.15) (1.29) (0.88) (0.48) PolicySurprise*Leverage 1.00* 8.29*** 13.97 17.60*** 5.19 6.44*** (0.60) (1.37) (8.91) (5.34) (3.70) (1.89) Leverage 0.08 –0.16** 1.45** –0.68*** –0.37* –0.33*** (0.07) (0.08) (0.62) (0.26) (0.20) (0.10) Constant –0.18 0.27*** –0.10 0.42*** 0.31*** 0.34*** (1.44) (0.02) (0.15) (0.06) (0.05) (0.03) Number of obs. 444,647 444,718 315,896 256,259 444,647 246,572 All regressions, except for (1), are instrumental variables regressions with instrument as specified in the column title. Regression 1 is a fixed effects panel regression that also includes industry dummies interacted with surprise. We include only firms that have positive debt and winsorize Leverage $$((\text{DLC}+\text{DLTT})/\text{AT})$$ at 1% level. Errors in parentheses are clustered at the firm level. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. The dependent variable is the stock returns of publicly listed companies. PolicySurprise is the unexpected federal funds target rate change on a scheduled FOMC meeting, as implied by federal funds futures. PolicySurprise is scaled so that positive surprises indicate expansionary shocks. Both returns and PolicySurprise are expressed in percentage points. HP and WW are dummy variables that are similarly defined for Hadlock-Pierce and Whited-Wu financial constraint measure and denote that a firm’s financial constraint measure is above the median in a given year. Unrated is a dummy variable equal to one if a firm does not have a long-term credit rating. Positive accruals is a dummy variable equal to one if the firm has positive accruals, as described in Table 1. Additional details are in Table 1. The sample includes firms traded in AMEX, NYSE, and NASDAQ. To ensure the liquidity of the stocks, penny stocks are dropped, following Amihud (2002), who defines penny stocks as stocks with a price less than $\$$5. In regressions including industry dummy interacted with PolicySurprise, the coefficient of PolicySurprise refers to the base industry which is nondurables. Table 9 Leverage and stock price sensitivity to monetary policy $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 * {\textit{PolicySurprise}}_t + \beta_2 * {\textit{Leverage}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Leverage}}_{\textit{it}-1} + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) (6) FE IV IV IV IV IV Variables UNRATED ACC>0 WW HP ALL PolicySurprise 1.99*** 0.96*** –0.44 –1.02 1.66* 1.58*** (0.38) (0.35) (2.15) (1.29) (0.88) (0.48) PolicySurprise*Leverage 1.00* 8.29*** 13.97 17.60*** 5.19 6.44*** (0.60) (1.37) (8.91) (5.34) (3.70) (1.89) Leverage 0.08 –0.16** 1.45** –0.68*** –0.37* –0.33*** (0.07) (0.08) (0.62) (0.26) (0.20) (0.10) Constant –0.18 0.27*** –0.10 0.42*** 0.31*** 0.34*** (1.44) (0.02) (0.15) (0.06) (0.05) (0.03) Number of obs. 444,647 444,718 315,896 256,259 444,647 246,572 $$\textit{return}_{\textit{it}} = \beta_0 + \beta_1 * {\textit{PolicySurprise}}_t + \beta_2 * {\textit{Leverage}}_{\textit{it}-1} + \delta * {\textit{PolicySurprise}}_t * {\textit{Leverage}}_{\textit{it}-1} + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) (6) FE IV IV IV IV IV Variables UNRATED ACC>0 WW HP ALL PolicySurprise 1.99*** 0.96*** –0.44 –1.02 1.66* 1.58*** (0.38) (0.35) (2.15) (1.29) (0.88) (0.48) PolicySurprise*Leverage 1.00* 8.29*** 13.97 17.60*** 5.19 6.44*** (0.60) (1.37) (8.91) (5.34) (3.70) (1.89) Leverage 0.08 –0.16** 1.45** –0.68*** –0.37* –0.33*** (0.07) (0.08) (0.62) (0.26) (0.20) (0.10) Constant –0.18 0.27*** –0.10 0.42*** 0.31*** 0.34*** (1.44) (0.02) (0.15) (0.06) (0.05) (0.03) Number of obs. 444,647 444,718 315,896 256,259 444,647 246,572 All regressions, except for (1), are instrumental variables regressions with instrument as specified in the column title. Regression 1 is a fixed effects panel regression that also includes industry dummies interacted with surprise. We include only firms that have positive debt and winsorize Leverage $$((\text{DLC}+\text{DLTT})/\text{AT})$$ at 1% level. Errors in parentheses are clustered at the firm level. $$^{***}$$p<.01, $$^{**}$$p<.05, $$^*$$p<.1. The dependent variable is the stock returns of publicly listed companies. PolicySurprise is the unexpected federal funds target rate change on a scheduled FOMC meeting, as implied by federal funds futures. PolicySurprise is scaled so that positive surprises indicate expansionary shocks. Both returns and PolicySurprise are expressed in percentage points. HP and WW are dummy variables that are similarly defined for Hadlock-Pierce and Whited-Wu financial constraint measure and denote that a firm’s financial constraint measure is above the median in a given year. Unrated is a dummy variable equal to one if a firm does not have a long-term credit rating. Positive accruals is a dummy variable equal to one if the firm has positive accruals, as described in Table 1. Additional details are in Table 1. The sample includes firms traded in AMEX, NYSE, and NASDAQ. To ensure the liquidity of the stocks, penny stocks are dropped, following Amihud (2002), who defines penny stocks as stocks with a price less than $\$$5. In regressions including industry dummy interacted with PolicySurprise, the coefficient of PolicySurprise refers to the base industry which is nondurables. As a final evidence of the leverage channel in the context of the Enron scandal, Table 10 reexamines the robustness of the pattern in Figure 2 and shows that this pattern conforms with the stock price sensitivity results in Section 2. This table uses a regression with firm and year fixed effects and the interaction of year fixed effects with a dummy variable that captures whether the firm was a client of Arthur Andersen ($$AAClient$$). The top panel of the first column shows that Arthur Andersen clients’ leverage was similar to other firms’ leverage before the scandal but that it declined relative to other firms’ leverage after 2001. This relative decline reached 16% by 2005 and reverted back thereafter, a timing that overlaps with the Supreme Court’s reversal of Arthur Andersen’s conviction. Columns 2 to 5 include additional control variables widely used in the corporate finance literature, taken from Lemmon, Roberts, and Zender (2008) and Frank and Goyal (2009). Since these two papers omit firm fixed effects in their main analysis, Column 2 provides the results without firm fixed effects to show that all the control variables, which are described further in the table notes, have the same signs as in these papers. Column 3 shows that adding firm fixed effects changes the coefficient of the industry median leverage significantly, a finding that is also consistent with these papers. More importantly, the Arthur Andersen effect remains similar across the first three columns. Table 10 The effect of the ENRON scandal of 2001 on the leverage of Arthur Andersen clients versus those of other firms (base year=2001) $$\log({\textit{Leverage}}_{\textit{it}}) = \beta_0 + {\textit{AAClient}}_i * (\beta_1 * d_{i,1997} + \beta_2 d_{i,1998} + \dotsc) + \gamma_1 * d_{i,1997} + \gamma_2 * d_{i,1998} + \dotsb + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) Variables Without controls With controls With controls RATED UNRATED AAClient*(Year=1997) 0.05 0.05 0.08 0.02 0.09 (0.08) (0.10) (0.11) (0.11) (0.17) AAClient*(Year=1998) 0.08 0.09 0.10 0.14 0.07 (0.07) (0.08) (0.09) (0.10) (0.15) AAClient*(Year=1999) 0.07 0.03 0.03 0.09 –0.05 (0.07) (0.08) (0.09) (0.10) (0.14) AAClient*(Year=2000) 0.02 0.03 0.02 0.10 –0.01 (0.07) (0.06) (0.07) (0.09) (0.10) AAClient*(Year=2002) –0.07 –0.09 –0.08 0.06 –0.18* (0.07) (0.06) (0.06) (0.05) (0.10) AAClient*(Year=2003) –0.13* –0.18** –0.19** 0.05 –0.34*** (0.07) (0.08) (0.08) (0.07) (0.13) AAClient*(Year=2004) –0.13* –0.15* –0.17** 0.10 –0.35** (0.07) (0.08) (0.08) (0.07) (0.14) AAClient*(Year=2005) –0.16** –0.20** –0.22** 0.01 –0.42*** (0.07) (0.09) (0.09) (0.08) (0.16) AAClient*(Year=2006) –0.05 –0.05 –0.08 –0.04 –0.16 (0.07) (0.09) (0.09) (0.08) (0.17) AAClient*(Year=2007) –0.10 –0.04 –0.04 0.08 –0.25 (0.08) (0.10) (0.11) (0.11) (0.19) Year=1997 –0.17*** 0.06* –0.02 –0.02 0.01 (0.03) (0.04) (0.04) (0.04) (0.06) Year=1998 –0.00 0.11*** 0.06 0.03 0.10* (0.03) (0.04) (0.04) (0.04) (0.05) Year=1999 0.03 0.08** 0.07** 0.03 0.10** (0.03) (0.03) (0.03) (0.03) (0.05) Year=2000 –0.01 0.07** 0.07** –0.00 0.11*** (0.03) (0.03) (0.03) (0.02) (0.04) Year=2002 0.02 0.02 0.01 –0.01 0.01 (0.03) (0.02) (0.02) (0.02) (0.04) Year=2003 –0.07** –0.08** –0.08** –0.07** –0.11** (0.03) (0.03) (0.03) (0.03) (0.05) Year=2004 –0.12*** –0.09** –0.11*** –0.16*** –0.12** (0.03) (0.04) (0.04) (0.04) (0.06) Year=2005 –0.13*** –0.08** –0.12*** –0.17*** –0.12* (0.03) (0.04) (0.04) (0.04) (0.06) Year=2006 –0.16*** –0.08* –0.15*** –0.15*** –0.15** (0.03) (0.04) (0.05) (0.04) (0.07) Year=2007 –0.08*** –0.05 –0.13** –0.13*** –0.13 (0.03) (0.05) (0.05) (0.04) (0.08) $$\log({\textit{Leverage}}_{\textit{it}}) = \beta_0 + {\textit{AAClient}}_i * (\beta_1 * d_{i,1997} + \beta_2 d_{i,1998} + \dotsc) + \gamma_1 * d_{i,1997} + \gamma_2 * d_{i,1998} + \dotsb + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) Variables Without controls With controls With controls RATED UNRATED AAClient*(Year=1997) 0.05 0.05 0.08 0.02 0.09 (0.08) (0.10) (0.11) (0.11) (0.17) AAClient*(Year=1998) 0.08 0.09 0.10 0.14 0.07 (0.07) (0.08) (0.09) (0.10) (0.15) AAClient*(Year=1999) 0.07 0.03 0.03 0.09 –0.05 (0.07) (0.08) (0.09) (0.10) (0.14) AAClient*(Year=2000) 0.02 0.03 0.02 0.10 –0.01 (0.07) (0.06) (0.07) (0.09) (0.10) AAClient*(Year=2002) –0.07 –0.09 –0.08 0.06 –0.18* (0.07) (0.06) (0.06) (0.05) (0.10) AAClient*(Year=2003) –0.13* –0.18** –0.19** 0.05 –0.34*** (0.07) (0.08) (0.08) (0.07) (0.13) AAClient*(Year=2004) –0.13* –0.15* –0.17** 0.10 –0.35** (0.07) (0.08) (0.08) (0.07) (0.14) AAClient*(Year=2005) –0.16** –0.20** –0.22** 0.01 –0.42*** (0.07) (0.09) (0.09) (0.08) (0.16) AAClient*(Year=2006) –0.05 –0.05 –0.08 –0.04 –0.16 (0.07) (0.09) (0.09) (0.08) (0.17) AAClient*(Year=2007) –0.10 –0.04 –0.04 0.08 –0.25 (0.08) (0.10) (0.11) (0.11) (0.19) Year=1997 –0.17*** 0.06* –0.02 –0.02 0.01 (0.03) (0.04) (0.04) (0.04) (0.06) Year=1998 –0.00 0.11*** 0.06 0.03 0.10* (0.03) (0.04) (0.04) (0.04) (0.05) Year=1999 0.03 0.08** 0.07** 0.03 0.10** (0.03) (0.03) (0.03) (0.03) (0.05) Year=2000 –0.01 0.07** 0.07** –0.00 0.11*** (0.03) (0.03) (0.03) (0.02) (0.04) Year=2002 0.02 0.02 0.01 –0.01 0.01 (0.03) (0.02) (0.02) (0.02) (0.04) Year=2003 –0.07** –0.08** –0.08** –0.07** –0.11** (0.03) (0.03) (0.03) (0.03) (0.05) Year=2004 –0.12*** –0.09** –0.11*** –0.16*** –0.12** (0.03) (0.04) (0.04) (0.04) (0.06) Year=2005 –0.13*** –0.08** –0.12*** –0.17*** –0.12* (0.03) (0.04) (0.04) (0.04) (0.06) Year=2006 –0.16*** –0.08* –0.15*** –0.15*** –0.15** (0.03) (0.04) (0.05) (0.04) (0.07) Year=2007 –0.08*** –0.05 –0.13** –0.13*** –0.13 (0.03) (0.05) (0.05) (0.04) (0.08) Market-to-book assets –0.04*** –0.04*** –0.02 –0.04*** (0.01) (0.01) (0.02) (0.01) Log(Sales) 0.10*** 0.06** –0.03 0.05 (0.01) (0.03) (0.03) (0.03) Profitability –0.33*** –0.35*** –0.79*** –0.31*** (0.07) (0.10) (0.17) (0.09) Tangibility 1.52*** 1.08*** 0.17 1.39*** (0.08) (0.17) (0.15) (0.21) Industry median leverage 0.39*** 0.11 0.18** 0.05 (0.05) (0.07) (0.08) (0.11) Market-to-book assets –0.04*** –0.04*** –0.02 –0.04*** (0.01) (0.01) (0.02) (0.01) Log(Sales) 0.10*** 0.06** –0.03 0.05 (0.01) (0.03) (0.03) (0.03) Profitability –0.33*** –0.35*** –0.79*** –0.31*** (0.07) (0.10) (0.17) (0.09) Tangibility 1.52*** 1.08*** 0.17 1.39*** (0.08) (0.17) (0.15) (0.21) Industry median leverage 0.39*** 0.11 0.18** 0.05 (0.05) (0.07) (0.08) (0.11) Number of observations 32,056 20,309 20,309 6,876 13,433 Number of GVKEYs 7,381 5,089 5,089 1,581 3,871 Number of observations 32,056 20,309 20,309 6,876 13,433 Number of GVKEYs 7,381 5,089 5,089 1,581 3,871 All regressions, except for (2), are firm fixed effects regressions. Regression (2) is a random-effects regression. In regressions (3)–(5), errors are clustered at firm level. $$^***$$p<.01, $$^**$$p<.05, $$^*$$p<.1. All balance sheet variables are from Compustat. The dependent variable is logarithm of book leverage defined as the sum of short-term debt (Compustat Item DLC) and long-term debt (DLTT) divided by total assets (AT). AAClient is a dummy that captures whether the firm was a client of Arthur Andersen in 2001. Market-to-book assets is total assets plus the market value of equity minus the book value of equity divided by total assets, where book value of equity is equal to the sum of common equity (CEQ) and deferred taxes (TXDITC). log(Sales) is the logarithm of sales (SALE). Profitability is operating income (OIBDP) divided by total assets. Tangibility is net plant, property, and equipment (PPENT) divided by total assets. The sample includes only firms with December fiscal year-end, so every observation in a given fiscal year belongs to the same calendar period to allow comparability across firms. Table 10 The effect of the ENRON scandal of 2001 on the leverage of Arthur Andersen clients versus those of other firms (base year=2001) $$\log({\textit{Leverage}}_{\textit{it}}) = \beta_0 + {\textit{AAClient}}_i * (\beta_1 * d_{i,1997} + \beta_2 d_{i,1998} + \dotsc) + \gamma_1 * d_{i,1997} + \gamma_2 * d_{i,1998} + \dotsb + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) Variables Without controls With controls With controls RATED UNRATED AAClient*(Year=1997) 0.05 0.05 0.08 0.02 0.09 (0.08) (0.10) (0.11) (0.11) (0.17) AAClient*(Year=1998) 0.08 0.09 0.10 0.14 0.07 (0.07) (0.08) (0.09) (0.10) (0.15) AAClient*(Year=1999) 0.07 0.03 0.03 0.09 –0.05 (0.07) (0.08) (0.09) (0.10) (0.14) AAClient*(Year=2000) 0.02 0.03 0.02 0.10 –0.01 (0.07) (0.06) (0.07) (0.09) (0.10) AAClient*(Year=2002) –0.07 –0.09 –0.08 0.06 –0.18* (0.07) (0.06) (0.06) (0.05) (0.10) AAClient*(Year=2003) –0.13* –0.18** –0.19** 0.05 –0.34*** (0.07) (0.08) (0.08) (0.07) (0.13) AAClient*(Year=2004) –0.13* –0.15* –0.17** 0.10 –0.35** (0.07) (0.08) (0.08) (0.07) (0.14) AAClient*(Year=2005) –0.16** –0.20** –0.22** 0.01 –0.42*** (0.07) (0.09) (0.09) (0.08) (0.16) AAClient*(Year=2006) –0.05 –0.05 –0.08 –0.04 –0.16 (0.07) (0.09) (0.09) (0.08) (0.17) AAClient*(Year=2007) –0.10 –0.04 –0.04 0.08 –0.25 (0.08) (0.10) (0.11) (0.11) (0.19) Year=1997 –0.17*** 0.06* –0.02 –0.02 0.01 (0.03) (0.04) (0.04) (0.04) (0.06) Year=1998 –0.00 0.11*** 0.06 0.03 0.10* (0.03) (0.04) (0.04) (0.04) (0.05) Year=1999 0.03 0.08** 0.07** 0.03 0.10** (0.03) (0.03) (0.03) (0.03) (0.05) Year=2000 –0.01 0.07** 0.07** –0.00 0.11*** (0.03) (0.03) (0.03) (0.02) (0.04) Year=2002 0.02 0.02 0.01 –0.01 0.01 (0.03) (0.02) (0.02) (0.02) (0.04) Year=2003 –0.07** –0.08** –0.08** –0.07** –0.11** (0.03) (0.03) (0.03) (0.03) (0.05) Year=2004 –0.12*** –0.09** –0.11*** –0.16*** –0.12** (0.03) (0.04) (0.04) (0.04) (0.06) Year=2005 –0.13*** –0.08** –0.12*** –0.17*** –0.12* (0.03) (0.04) (0.04) (0.04) (0.06) Year=2006 –0.16*** –0.08* –0.15*** –0.15*** –0.15** (0.03) (0.04) (0.05) (0.04) (0.07) Year=2007 –0.08*** –0.05 –0.13** –0.13*** –0.13 (0.03) (0.05) (0.05) (0.04) (0.08) $$\log({\textit{Leverage}}_{\textit{it}}) = \beta_0 + {\textit{AAClient}}_i * (\beta_1 * d_{i,1997} + \beta_2 d_{i,1998} + \dotsc) + \gamma_1 * d_{i,1997} + \gamma_2 * d_{i,1998} + \dotsb + {e}_{\textit{it}}$$ (1) (2) (3) (4) (5) Variables Without controls With controls With controls RATED UNRATED AAClient*(Year=1997) 0.05 0.05 0.08 0.02 0.09 (0.08) (0.10) (0.11) (0.11) (0.17) AAClient*(Year=1998) 0.08 0.09 0.10 0.14 0.07 (0.07) (0.08) (0.09) (0.10) (0.15) AAClient*(Year=1999) 0.07 0.03 0.03 0.09 –0.05 (0.07) (0.08) (0.09) (0.10) (0.14) AAClient*(Year=2000) 0.02 0.03 0.02 0.10 –0.01 (0.07) (0.06) (0.07) (0.09) (0.10) AAClient*(Year=2002) –0.07 –0.09 –0.08 0.06 –0.18* (0.07) (0.06) (0.06) (0.05) (0.10) AAClient*(Year=2003) –0.13* –0.18** –0.19** 0.05 –0.34*** (0.07) (0.08) (0.08) (0.07) (0.13) AAClient*(Year=2004) –0.13* –0.15* –0.17** 0.10 –0.35** (0.07) (0.08) (0.08) (0.07) (0.14) AAClient*(Year=2005) –0.16** –0.20** –0.22** 0.01 –0.42*** (0.07) (0.09) (0.09) (0.08) (0.16) AAClient*(Year=2006) –0.05 –0.05 –0.08 –0.04 –0.16 (0.07) (0.09) (0.09) (0.08) (0.17) AAClient*(Year=2007) –0.10 –0.04 –0.04 0.08 –0.25 (0.08) (0.10) (0.11) (0.11) (0.19) Year=1997 –0.17*** 0.06* –0.02 –0.02 0.01 (0.03) (0.04) (0.04) (0.04) (0.06) Year=1998 –0.00 0.11*** 0.06 0.03 0.10* (0.03) (0.04) (0.04) (0.04) (0.05) Year=1999 0.03 0.08** 0.07** 0.03 0.10** (0.03) (0.03) (0.03) (0.03) (0.05) Year=2000 –0.01 0.07** 0.07** –0.00 0.11*** (0.03) (0.03) (0.03) (0.02) (0.04) Year=2002 0.02 0.02 0.01 –0.01 0.01 (0.03) (0.02) (0.02) (0.02) (0.04) Year=2003 –0.07** –0.08** –0.08** –0.07** –0.11** (0.03) (0.03) (0.03) (0.03) (0.05) Year=2004 –0.12*** –0.09** –0.11*** –0.16*** –0.12** (0.03) (0.04) (0.04) (0.04) (0.06) Year=2005 –0.13*** –0.08** –0.12*** –0.17*** –0.12* (0.03) (0.04) (0.04) (0.04) (0.06) Year=2006 –0.16*** –0.08* –0.15*** –0.15*** –0.15** (0.03) (0.04) (0.05) (0.04) (0.07) Year=2007 –0.08*** –0.05 –0.13** –0.13*** –0.13 (0.03) (0.05) (0.05) (0.04) (0.08) Market-to-book assets –0.04*** –0.04*** –0.02 –0.04*** (0.01) (0.01) (0.02) (0.01) Log(Sales) 0.10*** 0.06** –0.03 0.05 (0.01) (0.03) (0.03) (0.03) Profitability –0.33*** –0.35*** –0.79*** –0.31*** (0.07) (0.10) (0.17) (0.09) Tangibility 1.52*** 1.08*** 0.17 1.39*** (0.08) (0.17) (0.15) (0.21) Industry median leverage 0.39*** 0.11 0.18** 0.05 (0.05) (0.07) (0.08) (0.11) Market-to-book assets –0.04*** –0.04*** –0.02 –0.04*** (0.01) (0.01) (0.02) (0.01) Log(Sales) 0.10*** 0.06** –0.03 0.05 (0.01) (0.03) (0.03) (0.03) Profitability –0.33*** –0.35*** –0.79*** –0.31*** (0.07) (0.10) (0.17) (0.09) Tangibility 1.52*** 1.08*** 0.17 1.39*** (0.08) (0.17) (0.15) (0.21) Industry median leverage 0.39*** 0.11 0.18** 0.05 (0.05) (0.07) (0.08) (0.11) Number of observations 32,056 20,309 20,309 6,876 13,433 Number of GVKEYs 7,381 5,089 5,089 1,581 3,871 Number of observations 32,056 20,309 20,309 6,876 13,433 Number of GVKEYs 7,381 5,089 5,089 1,581 3,871 All regressions, except for (2), are firm fixed effects regressions. Regression (2) is a random-effects regression. In regressions (3)–(5), errors are clustered at firm level. $$^***$$p<.01, $$^**$$p<.05, $$^*$$p<.1. All balance sheet variables are from Compustat. The dependent variable is logarithm of book leverage defined as the sum of short-term debt (Compustat Item DLC) and long-term debt (DLTT) divided by total assets (AT). AAClient is a dummy that captures whether the firm was a client of Arthur Andersen in 2001. Market-to-book assets is total assets plus the market value of equity minus the book value of equity divided by total assets, where book value of equity is equal to the sum of common equity (CEQ) and deferred taxes (TXDITC). log(Sales) is the logarithm of sales (SALE). Profitability is operating income (OIBDP) divided by total assets. Tangibility is net plant, property, and equipment (PPENT) divided by total assets. The sample includes only firms with December fiscal year-end, so every observation in a given fiscal year belongs to the same calendar period to allow comparability across firms. Columns 4 and 5 show that the measured effect stems primarily from the sample of unrated firms, consistent with the raw data plots in Figure 3. This finding conforms with the result in Section 2 that the effect of the scandal on the policy sensitivity of stocks is concentrated among unrated firms. Moreover, Column 5 confirms that the relative post-2001 decline in Arthur Andersen clients’ leverage cannot be attributed to the continuation of a time trend from before 2001. If anything, among unrated firms, Arthur Andersen clients’ leverage has increased slightly relative to other firms’ leverage over the period from 1999 to 2001, but this increase ($$0.05$$) is economically and statistically insignificant ($$t=0.35$$). Overall, these results suggest that Arthur Andersen clients’ debt has decreased by about 20% to 40% relative to other firms. This gives a lower bound for the magnitude of the change in the monetary policy responsiveness of Arthur Andersen clients because there was no certainty that things would begin to return to normal after only four years. Indeed, if the Supreme Court had not reversed Arthur Andersen’s conviction in 2005, things might have been much worse for these firms.24 Given that a 10-bp surprise reduction in the federal funds target rate has increased the stock prices by about 1 percentage point in 2001, we should expect that Arthur Andersen firms should react by at least about 0.2 to 0.4 percentage points less toward the end of the scandal, which is not too far off from the numbers found in the previous section (about 0.5 percentage points). Figure 3 View largeDownload slide Average leverage of Arthur Andersen clients and other firms by rating status The top (bottom) panel presents the average leverage of firms forming the sample of rated (unrated) firms in the leverage regressions. The solid lines are for firms that were clients of Arthur Andersen (AA); the dash-dot lines are for other firms. Figure 3 View largeDownload slide Average leverage of Arthur Andersen clients and other firms by rating status The top (bottom) panel presents the average leverage of firms forming the sample of rated (unrated) firms in the leverage regressions. The solid lines are for firms that were clients of Arthur Andersen (AA); the dash-dot lines are for other firms. Although there may be channels other than leverage that explain the main results, these other channels are not as direct as the leverage channel or they are more difficult to test due to data unavailability. One promising channel is related to the source of debt financing. For example, firms suffering from greater information frictions may end up using more bank debt because banks can deal with information frictions better than other creditors. The increased reliance on banks then can provide a cushion against fluctuations in interest rates due to an imperfect pass-through of interest rates to loan rates or because bank-firm relationships enable firms to benefit from some degree of insurance against changes in credit availability (Puri, Rocholl, and Steffen (2013)). However, in contrast with this argument, bank-financed debt usually has a floating interest rate, which makes stock prices more responsive to monetary policy (Ippolito, Ozdagli, and Perez (2017)). The use of floating versus fixed rate debt brings up another related channel that may explain our results: less financially constrained firms may be able to borrow more variable rate debt, which makes them more responsive through the floating rate channel discussed in Ippolito, Ozdagli, and Perez (2017). Still, the results of Ippolito, Ozdagli, and Perez (2017, table A4) suggest that the stocks of firms with larger size or smaller Handlock-Pierce index, which are usually considered less constrained in the literature, remain more responsive to monetary policy even after controlling for variable rate debt (which is usually bank debt). For the pre-Enron-scandal period, reliable data on variable rate debt is not publicly available in a commercial database. Nevertheless, the Capital IQ dataset of debt structure suggests that the ratio of the variable rate debt to total assets of Arthur Andersen ex-clients changed little relative to other firms since 2002 (about 0.12 and 0.10 in 2002 and 0.09 and 0.07 in 2005). As a result, while floating rate channel may be another factor contributing to the change in the monetary policy sensitivity of stock prices, it does not seem to fully explain the cross-sectional differences in policy sensitivity generated by financial frictions. 4. Conclusion This paper studies the relationship between information frictions and the reaction of stock prices to monetary policy shocks. The empirical analysis shows that information frictions play an important role to explain the cross-sectional differences in monetary policy sensitivity of stock prices. In particular, the stock prices of firms subject to greater information frictions react less to monetary policy. The paper presents strong empirical evidence based on both broad proxies and the differential effect of the Enron/Arthur Andersen scandal on other Arthur Andersen clients’ sensitivity to monetary policy after the scandal. The results are robust to several identification methods including difference-in-differences, intraday returns, instrumental variables, panel data analysis, and a placebo experiment. The analysis of a longer panel also reveals that financial restatements reduce the effect of monetary policy on stock prices of restating firms, consistent with the findings from the Enron scandal. An important message of the paper is that any empirical analysis of financial constraints, monetary policy, and stock prices should be based on a clear definition of the source of financial constraint being examined, which is, in our case, information frictions. Although an important underlying element of most financial constraints, information frictions are certainly not the only source of financial constraints. It would also be interesting to analyze and test the implications of other types of financial constraints, such as those that originate from the structure and regulations of the financial system, on the relationship between stock prices and monetary policy. Continuing this ambitious research agenda is left for future work. I thank Robin Greenwood (the editor) and two anonymous referees for their constructive comments and suggestions. I also thank Gabriel Chodorow-Reich, Adam Guren, Jens Hilscher, Peter Karadi, Anil Kashyap, Ralf Meisenzahl, Fabio Schiantarelli, and, in particular, Victoria Ivashina and Joe Peek for their helpful feedback. Seminar participants at Boston University, Green Line Macro Meeting at Boston College, Federal Reserve Bank of Boston, Federal Reserve Bank of Dallas, University of Wisconsin Business School, the Society for Economic Dynamics meeting in Toronto, and the North American Econometric Society meeting at the University of Minnesota provided useful remarks at various stages of the paper. Suzanne Lorant, Jianlin Wang, and Yifan Yu provided excellent editorial and research assistance. The views expressed in this paper do not necessarily reflect those of the Federal Reserve Bank of Boston or the Federal Reserve System. Supplementary data can be found on The Review of Financial Studies Web site. Appendix A. The Model and an Empirical Prediction This section shows how the monetary policy sensitivity of a firm’s market value of equity can be lower if the firm faces greater information frictions. For this purpose, I closely follow the popular framework in Bernanke, Gertler, and Gilchrist (1999), appendix A in particular. The firm with net worth (book equity) $$N$$ chooses the value of its capital, $$QK$$, where $$Q$$ is the price of capital. It borrows $$QK-N$$ in exchange of the face value of debt, $$B$$. After these decisions, the firm’s profitability, $$w$$, is realized. The firm maximizes shareholder value \begin{equation} V=\max_{K,B}\frac{1}{R}E\left( wQK-B\right) ^{+}, \end{equation} (A1) subject to the incentive compatibility constraint of the lender that makes the lender indifferent between investing $$QK-N$$ at the risk-free rate, $$R$$, or lending it to the firm. This constraint is given by the equation \begin{equation} R\left( QK-N\right) =E\left( \mathbb{I}_{wQK\geq B}B+\mathbb{I}_{\textit{wQK}<B}\left( 1-\mu \right) wQK\right) , \end{equation} (A2) where $$R$$ is the gross risk-free rate, $$\mu $$ is the auditing cost that the lender incurs if the lender does not pay the promised amount $$B$$, and $$\mathbb{I}$$ denotes the indicator function that is equal to one if the corresponding condition is satisfied and zero otherwise. $$E\left( \cdot \right) $$ is the expectation taken over $$w$$, like in Bernanke, Gertler, and Gilchrist (1999).25 The object of interest is $$d\ln V/d\ln R$$ because the interest rate is set by the monetary authority in this model. Defining $$v\equiv V/N,k\equiv QK/N$$, and $$\bar{w}\equiv B/QK$$, we can rewrite the firm’s problem as \begin{equation} v=\max_{k,\bar{w}}\frac{1}{R}E\left( w-\bar{w}\right) ^{+}k, \label{eq: V/N} \end{equation} (A3) subject to \begin{equation} R\left( k-1\right) =E\left( \mathbb{I}_{w\geq \bar{w}}\bar{w}+\mathbb{I}_{w<\bar{w}}\left( 1-\mu \right) w\right) k. \label{eq: constr/N} \end{equation} (A4) The analysis focuses on how the percentage change in stock prices in response to a change in the risk-free rate varies with auditing costs, the main source of information friction in this model. Because net worth, $$N$$, is a state variable independent of the interest rate, as it is the case in BGG (1999), one can directly work with $$v$$, instead of $$V$$, which simplifies the problem. The Online Appendix presents an extension by which $$N$$ is not a state variable, and the results in this section continue to hold. The first proposition below shows that firms using more debt are more responsive to monetary policy shocks. Proposition 1. The sensitivity of a firm’s stock value to monetary policy shocks, $$\left\vert d\ln V/d\ln R\right\vert $$, increases as the ratio of total capital tonet worth, $$QK/N$$, increases. Proof. Solving constraint (A4) for $$k$$ and plugging the solution into the objective function (A3) results in an unconstrained problem in $$\bar{w}$$. Then, using the envelope theorem, we obtain $$d\ln V/d\ln R=-R/\left[ R-E\left( \mathbb{I}_{w\geq \bar{w}}\bar{w}+\mathbb{I}_{w<\bar{w}}\left( 1-\mu \right) w\right) \right] $$. Using constraint (A4), this reduces to $$d\ln V/d\ln R=-k<0$$. Moreover, $$\left\vert d\ln V/d\ln R\right\vert $$increases in $$k\equiv QK/N$$. ■ Intuitively, monetary policy affects a firm’s behavior by changing its cost of debt. Therefore, a firm that relies more on debt will be more affected by a monetary policy shock. The next proposition establishes that firms subject to greater auditing costs, the main source of information friction in this model, use less debt. Proposition 2. Let $$f\left( w\right) $$ and $$F\left( w\right) $$ denote the pdf and cdf of the firm’s productivity, $$h\left( w\right) \equiv f\left( w\right) /\left( 1-F\left( w\right) \right) $$ denote the hazard rate, and let $$\bar{w}h\left( \bar{w}\right) $$ be increasing in $$\bar{w}$$. Then, the ratio of total capital to net worth, $$QK/N$$, islower for firms with greater auditing costs, $$\mu $$. Proof. See Appendix B. ■ Intuitively, firms with higher auditing costs use less debt than they otherwise would because they have to pay a higher cost of debt. The assumption regarding the hazard rate is imposed by Bernanke, Gertler, and Gilchrist (1999) to guarantee a nonrationing outcome, which is particularly realistic for the publicly listed firms being studied. Additional details can be found in appendix A.1 of Bernanke, Gertler, and Gilchrist (1999). Together, these two propositions yield the conclusion that firms with higher auditing costs rely less on debt than they otherwise would, a fact that makes their stock prices less sensitive to monetary policy shocks. Therefore, the two propositions lead to the following corollary, which is the main hypothesis of this paper. Corollary 1. Firms with greater information frictions have lower stock price sensitivity to monetary policy. While this section focuses on a partial equilibrium framework, interest rates in general equilibrium can affect the price, $$Q$$, if thesupply of capital goods is not perfectly elastic, in which case $$Q\equiv Q(R)$$. Similarly, the price of capital can also be affectedby the cross-sectional distribution of auditing costs, say $$M\left( \mu \right) $$, in which case $$Q\equiv Q(R,M\left( \mu \right) )$$.Nevertheless, the envelope theorem underlying the proof continues to hold even when $$Q\equiv Q(R,M\left( \mu \right) )$$, because choosing$$K$$ and $$QK$$ is the same from firm’s perspective, since the firm takes $$Q$$ as given. The Online Appendix presents a slight modificationof the BGG framework that results in a setting similar to that in Carlstrom and Fuerst (1997), where the reaction of $$Q$$ to $$R$$ generates amultiplier effect for the policy sensitivity of stock prices. However, because all firms face the same $$Q$$, this multiplier is thesame across the firms and hence the propositions in this section continue to hold.26 The simple theoretical analysis in this section omits other potential channels through which information frictions may reduce or amplify the effect of the channel studied here. For example, while a nonrationing equilibrium seems to be more realistic for the publicly listed companies being examined, credit rationing may still play a role. In this case, firms that previously had been rationed in the debt market may be able to borrow more, due to an accommodative monetary policy, and this situation could alleviate the effect of increased auditing costs. Similarly, if the managers maximize the value for existing shareholders and existing shareholders are passive investors, like in Majluf, and Myers (1984), firms subject to greater information frictions may prefer risk-free debt over equity. However, the final effect on leverage is not clear because internal equity (retained earnings) is still preferred over risk-free debt which may reduce leverage. Moreover, when the firm has to choose between risky debt and equity, the pecking order between debt and outside equity may break down (Noe (1988)). The Online Appendix presents an extension with choice of equity ($$N$$), and the results in this section continue to hold. Alternatively, firms that are harder to audit may have a greater perceived dispersion (uncertainty) in their profitability, $$w$$, which may further limit these firms’ borrowing, amplifying the effect of auditing costs.27 Ultimately, the net effect of information frictions is an empirical question that is addressed in this paper. At this point, it is also worth noting that auditing costs and uncertainty are two sides of the same coin. If auditing were costless, there would be no need for independent auditing of financial statements. When independent auditing fails, like in the case of the Arthur Andersen scandal, investors should pay the cost of generating their own information to the extent it is possible. However, if the information cannot be replicated at any cost then the effect would be an increase in the uncertainty about the firm’s prospects, which further reduces the firm’s ability to borrow. The theoretical result in this section may seem to contrast with the empirical studies on how the credit channel of monetary policy affects real variables, such as investment, as studied in Gertler and Gilchrist (1994) and Kashyap, Lamont, and Stein (1994). In general, these papers suggest that the real variables of financially constrained firms are more responsive to monetary policy. While it may be tempting to extend this argument to the sensitivity of stock prices to monetary policy, the analysis in this section urges caution against this temptation. In particular, the stock price is the value function of a maximization problem, where investment is the choice variable, and there is no theoretical necessity that the value function and the choice variable should move in the same direction in response to changes in the economic environment. Since the firm’s investment is determined by first-order conditions and its stock price is determined by the value function, the reaction of its investment involves a higher order derivative of the objective function than the reaction of its stock price. This theoretical analysis also illustrates the importance of studying stock prices to shed light on questions that may be more difficult to study otherwise. For example, firms’ capital structure or investment may respond to changes in the economic environment rather slowly and therefore by the time a measurable response is observed, the economic environment may have changed yet again, making identification particularly difficult. However, stock prices react to monetary policy immediately and allow high-frequency identification of both monetary policy surprises and their effects. Therefore, even though we might not observe large immediate changes in firms’ capital structure or investment, the anticipation of these future changes and the associated transmission mechanisms are immediately reflected in stock prices. From the perspective of the model, this corresponds to a simple relabeling of $$QK/N$$ as the beginning-of-future-period capital structure, which is anticipated today but will be realized in the future, where future can be a day, a month, a year, or longer depending on the speed of capital adjustment. As a result, although the model is very stylized, its implications extend beyond that. This potential of stock prices to address long-standing difficult questions involving monetary policy transmission has recently received greater attention.28 Appendix B. Proof of Proposition 2 Using constraint (A4), we can write \begin{equation} R\left( \frac{k-1}{k}\right) =\Gamma \left( \bar{w}\right) -\mu G\left( \bar{w}\right) , \end{equation} (A5) where \begin{equation} \Gamma \left( \bar{w}\right) -\mu G\left( \bar{w}\right) =\bar{w}+\int_{0}^{\bar{w}}\left( \left( 1-\mu \right) w-\bar{w}\right) dF\left( w\right) . \end{equation} (A6) It is clear that for a given value of $$\bar{w}$$, $$k$$ is decreasing in $$\mu $$. Moreover, Bernanke, Gertler, and Gilchrist (1999) show that \begin{eqnarray} \Gamma ^{\prime }\left( \bar{w}\right) -\mu G^{\prime }\left( \bar{w}\right) &=&1-F\left( \bar{w}\right) -\mu \bar{w}f\left( \bar{w}\right) \nonumber \\ &=&\left[ 1-F\left( \bar{w}\right) \right] \left[ 1-\mu \bar{w}h\left( \bar{w}\right) \right] >0 \end{eqnarray} (A7) in equilibrium if $$\bar{w}h\left( \bar{w}\right) $$ is increasing in $$\bar{w}$$. To summarize their argument, because $$\bar{w}h\left( \bar{w}\right) $$ is increasing in $$\bar{w},$$ there exists a $$\bar{w}^{\ast }$$ so that $$\Gamma ^{\prime }\left( \bar{w}\right) -\mu G^{\prime }\left( \bar{w}\right) \lesseqqgtr 0$$ if $$\bar{w}\gtreqqless \bar{w}^{\ast }$$, where $$\bar{w}^{\ast }$$ satisfies $$1-\mu \bar{w}^{\ast }h\left( \bar{w}^{\ast }\right) =0$$. Appendix A.1 of Bernanke, Gertler, and Gilchrist (1999) shows that $$\bar{w}>\bar{w}^{\ast }$$ cannot be an equilibrium. In particular, if the lender gives the firm $$QK-N$$, its expected payoff from this lending, $$E\left( \mathbb{I}_{wK\geq B}B+\mathbb{I}_{wK<B}\left( 1-\mu \right) wQK\right) =\left[ \Gamma \left( \bar{w}\right) -\mu G\left( \bar{w}\right) \right] QK$$, will decrease in the face value of debt, $$B=\bar{w}QK$$, for $$\bar{w}>\bar{w}^{\ast }$$ because $$\Gamma ^{\prime \prime }\left( \bar{w}\right) -\mu G^{\prime \prime }\left( \bar{w}\right) <0$$. Therefore, both the firm and the lender would benefit from a lower $$\bar{w}$$ when $$\bar{w}>\bar{w}^{\ast }$$. Hence, the equilibrium value of $$\bar{w}$$ cannot be in this region. As a result, the final step only needs to establish that $$d\bar{w}/d\mu <0$$. By substituting the incentive compatibility constraint (A4) of the lender into the objective function of the firm (A3), we obtain \begin{equation} v=\max_{\bar{w}}\frac{\int_{\bar{w}}^{\infty }\left( w-\bar{w}\right) dF\left( w\right) }{R-\left[ \bar{w}+\int_{0}^{\bar{w}}\left( \left( 1-\mu \right) w-\bar{w}\right) dF\left( \bar{w}\right) \right] }=\frac{P\left( \bar{w}\right) }{R-\left[ \Gamma \left( \bar{w}\right) -\mu G\left( \bar{w}\right) \right] }\text{,} \end{equation} (A8) which has the first order condition \begin{equation} {\it{\Omega}} \left( \bar{w},\mu \right) =P^{\prime }\left( \bar{w}\right) \left( R-\left[ \Gamma \left( \bar{w}\right) -\mu G\left( \bar{w}\right) \right] \right) +P\left( \bar{w}\right) \left[ \Gamma ^{\prime }\left( \bar{w}\right) -\mu G^{\prime }\left( \bar{w}\right) \right] =0, \end{equation} (A9) which should satisfy $$\partial {\it{\Omega}} \left( \bar{w},\mu \right) /\partial \bar{w}<0$$ at the equilibrium value of $$\bar{w}$$ because the second-order condition, $$d^{2}v/d\bar{w}^{2}<0$$, dictates that $${\it{\Omega}} \left( \bar{w}+\varepsilon ,\mu \right) >0$$ and $${\it{\Omega}} \left( \bar{w}-\varepsilon ,\mu \right) <0$$ for any positive value of $$\varepsilon $$ at the equilibrium value of $$\bar{w}$$. Full differentiation of both sides yields \begin{equation} \frac{\partial {\it{\Omega}} \left( \bar{w},\mu \right) }{\partial \bar{w}}\frac{d\bar{w}}{d\mu }=P\left( \bar{w}\right) G^{\prime }\left( \bar{w}\right) -P^{\prime }\left( \bar{w}\right) G\left( \bar{w}\right) \text{.} \end{equation} (A10) It is straightforward to show that the right side is positive which, combined with $$\partial {\it{\Omega}} \left( \bar{w},\mu \right) /\partial \bar{w}<0 $$, gives $$d\bar{w}/d\mu <0$$. References Agrawal, A., and Chadha. S. 2005. Corporate governance and accounting scandals. Journal of Law and Economics 48: 371– 4063. Google Scholar CrossRef Search ADS Amihud, Y. 2002. Illiquidity and stock returns: Cross-section and time-series effects. Journal of Financial Markets 5: 31– 56. Google Scholar CrossRef Search ADS Anantharaman, D., and Lee. Y. 2014. Managerial risk taking incentives and corporate pension policy. Journal of Financial Economics 111: 328– 51. Google Scholar CrossRef Search ADS Avramov, D., Chordia, T. Jostova, G. and Philipov. A. 2007. Momentum and credit rating. Journal of Finance 62: 2503– 20. Google Scholar CrossRef Search ADS Berk, J., Green, R. and Naik. V. 2004. Valuation and return dynamics of new ventures. Review of Financial Studies 17: 1– 35. Google Scholar CrossRef Search ADS Bernanke, B., and Gertler. M. 1995. Inside the black box: The credit channel of monetary policy transmission. Journal of Economic Perspectives 9: 27– 48. Google Scholar CrossRef Search ADS Bernanke, B., Gertler, M. and Gilchrist. S. 1999. The financial accelerator in a quantitative business cycle framework. In Handbook of macroeconomics , eds. Taylor J. B. and Woodford, M. 1341– 93. Amsterdam: North-Holland. Bernanke, B., and Kuttner. K. 2005. What explains the stock market’s reaction to Federal Reserve policy? Journal of Finance 60: 1221– 1257. Google Scholar CrossRef Search ADS Campbell, D., and Hulme. R. 2001. The winner-takes-all economy. McKinsey Quarterly , Winter. Campbell, J., Hilscher, J. and Szilagyi. J. 2008. In search of distress risk. Journal of Finance 63: 2899– 939. Google Scholar CrossRef Search ADS Cameron, A., and Trivedi. P. 2010. Microeconometrics using stata , Revised Edition. College Station, TX: Stata Press. Carlstrom, C., and Fuerst. T. 1997. Agency costs, net worth, and business fluctuations: A computable general equilibrium analysis. American Economic Review 87: 893– 910. Chaney, P., Jeter, D. and Shivakumar. L. 2004. Self-selection of auditors and audit pricing in private firms. Accounting Review 79: 51– 72. Google Scholar CrossRef Search ADS Chang, X., Dasgupta, S. and Hilary. G. 2009. The effect of auditor quality on financing decisions. Accounting Review 84: 1085– 117. Google Scholar CrossRef Search ADS Chava, S., and Hsu. A. 2015. Financial constraints, monetary policy shocks, and the cross-section of equity returns. Working Paper . Colla, P., Ippolito, F. and Li. K. 2013 Debt specialization. Journal of Finance 68: 2117– 41. Google Scholar CrossRef Search ADS Daniel, K., and Titman. S. 1997. Evidence on the characteristics of cross sectional variation in stock returns. Journal of Finance 52: 1– 33. Google Scholar CrossRef Search ADS Dyck, A., Morse, A. and Zingales. L. 2014. How pervasive is corporate fraud? Working Paper . Ehrmann, M., and Fratscher. M. 2004. Taking stock: Monetary policy transmission to equity markets. Journal of Money, Credit and Banking 36: 719– 37. Google Scholar CrossRef Search ADS Eisenberg, T., and Macey. J. 2004. Was Arthur Andersen different? An empirical examination of major accounting firm audits of large clients. Journal of Empirical Legal Studies 1: 263– 300. Google Scholar CrossRef Search ADS Evans, C. 1998. Real-time taylor rules and the federal funds futures market. Economic Perspectives 22: 44– 55. Fama, E., and French. K. 1992. The cross-section of expected stock returns. Journal of Finance 47: 427– 65. Google Scholar CrossRef Search ADS Fama, E., and French. K. 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33: 3– 56. Google Scholar CrossRef Search ADS Farre-Mensa, J., and Ljungqvist. A. 2015. Do measures of financial constraints measure financial constraints? Review of Financial Studies 29: 271– 308. Google Scholar CrossRef Search ADS Faulkender, M., and Petersen. M. 2006. Does the source of capital affect capital structure? Review of Financial Studies 19: 45– 79. Google Scholar CrossRef Search ADS Ferson, W., and Harvey. C. 1999. Conditioning variables and the cross section of stock returns. Journal of Finance 54: 1325– 60. Google Scholar CrossRef Search ADS Fisher, I. 1933. The debt-deflation theory of great depressions. Econometrica 1: 337– 57. Google Scholar CrossRef Search ADS Frank, M., and Goyal. V. 2009. Capital structure decisions: Which factors are reliably important? Financial Management 38: 1– 37. Google Scholar CrossRef Search ADS Gertler, M., and Gilchrist. S. 1994. Monetary policy, business cycles and the behavior of small manufacturing firms. Quarterly Journal of Economics 109: 309– 40. Google Scholar CrossRef Search ADS Gleason, C., Jenkins, N. and Johnson. W. 2008. The contagion effects of accounting restatements. Accounting Review 83: 83– 110. Google Scholar CrossRef Search ADS Gorodnichenko, Y., and Weber. M. 2016. Are sticky prices costly? Evidence from the stock market. American Economic Review 106: 165– 99. Google Scholar CrossRef Search ADS Graham, J., Hughson, E. and Zender. J. 1999. Market reactions to capital structure changes: Theory and evidence. Working Paper . Graham, J., Li, S. and Qiu. J. 2008. Corporate misreporting and bank loan contracting. Journal of Financial Economics 89: 44– 61. Google Scholar CrossRef Search ADS Federal Open Market Committee. 1995. Meeting of the Federal Open Market Committee. Meeting Minutes, January 31–February 1. Washington, DC: Federal Open Market Committee. https://www.federalreserve.gov/monetarypolicy/files/FOMC19950201meeting.pdf. Gürkaynak, R., Sack, B. and Swanson. E. 2007. Market-based measures of monetary policy expectations. Journal of Business and Economic Statistics 25: 201– 12. Google Scholar CrossRef Search ADS Hadlock, C., and Pierce. J. 2010. New evidence on measuring financial constraints: Moving beyond the KZ Index. Review of Financial Studies 23: 1909– 40. Google Scholar CrossRef Search ADS Holmström, B., and Tirole. J. 1997. Financial intermediation, loanable funds, and the real sector. Quarterly Journal of Economics 52: 663– 92. Google Scholar CrossRef Search ADS Ippolito, F., Ozdagli, A. and Perez. A. 2017. The transmission of monetary policy through bank lending: The floating rate channel. Working Paper , FEDS. Kaplan, S., and Zingales. L. 1997. Do investment-cash flow sensitivities provide useful measures of financing constraints? Quarterly Journal of Economics 115: 707– 12. Google Scholar CrossRef Search ADS Kashyap, A., Lamont, O. and Stein. J. 1994. Credit conditions and the cyclical behavior of inventories. Quarterly Journal of Economics 109: 565– 92. Google Scholar CrossRef Search ADS Kiyotaki, N., and Moore. J. 1997. Credit cycles. Journal of Political Economy 105: 211– 48. Google Scholar CrossRef Search ADS Kraus, A., and Litzenberger. R. 1973. A state-preference model of optimal financial leverage. Journal of Finance 28: 911– 22. Google Scholar CrossRef Search ADS Kuttner, K. 2001. Monetary policy surprises and interest rates: Evidence from the fed funds futures market. Journal of Monetary Economics 47: 523– 44. Google Scholar CrossRef Search ADS Lamont, O., Polk, C. and Saá-Requejo. J. 2001. Financial constraints and stock returns. Review of Financial Studies 14: 529– 54. Google Scholar CrossRef Search ADS Leary, M., and Roberts. M. 2010. The pecking order, debt capacity, and information asymmetry. Journal of Financial Economics 95: 332– 55. Google Scholar CrossRef Search ADS Lemmon, M., Roberts, M. and Zender. J. 2008. Back to the beginning: persistence and the cross-section of corporate capital structures. Journal of Finance 63: 1575– 608. Google Scholar CrossRef Search ADS Lucca, D., and Moench. E. 2015. The pre-FOMC announcement drift. Journal of Finance 70: 329– 71. Google Scholar CrossRef Search ADS Majluf, N. and Myers. S. 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13: 187– 221. Google Scholar CrossRef Search ADS McLean, B. 2001. Is Enron overpriced? Fortune , March 5. Modigliani, F. 1971. Consumer spending and monetary policy: The linkages. Federal Reserve Bank of Boston Conference Series 5: 9– 84. Noe, T. 1988. Capital structure and signaling game equilibria. Review of Financial Studies 1: 331– 56. Google Scholar CrossRef Search ADS Ozdagli, A. 2012. Financial leverage, corporate investment, and stock returns. Review of Financial Studies 25: 1033– 69. Google Scholar CrossRef Search ADS Perez-Quiros, G., and Timmermann. A. 2000. Firm size and cyclical variations in stock returns. Journal of Finance 55: 1229– 62. Google Scholar CrossRef Search ADS Piazzesi, M., and Swanson. E. 2008. Futures prices as risk-adjusted forecasts of monetary policy. Journal of Monetary Economics 55: 677– 91. Google Scholar CrossRef Search ADS Polk, C., and Sapienza. P. 2009. The stock market and corporate investment: A test of catering theory. Review of Financial Studies 22: 187– 217. Google Scholar CrossRef Search ADS Puri, M., Rocholl, J. and Steffen. S. 2013. What kinds of bank-client relationships matter in reducing loan defaults and why? Working Paper . Roberts, M., and Whited. T. 2013. Endogeneity in empirical corporate finance. In Handbook of the economics of finance , eds. Constantinides, G. M. Harris, M. and Stulz, R. M. 493– 572. Amsterdam: North Holland. Google Scholar CrossRef Search ADS Sellin, P. 2001. Monetary policy and the stock market: Theory and empirical evidence. Journal of Economic Surveys 15: 491– 541. Google Scholar CrossRef Search ADS Sloan, R. 1996. Do stock prices fully reflect information in accruals and cash flows? Accounting Review Spring: 289– 316. Sufi, A. 2007. Information asymmetry and financing arrangements: Evidence from syndicated loans. Journal of Finance 62: 629– 68. Google Scholar CrossRef Search ADS Tobin, J. 1969. A General equilibrium approach to monetary theory. Journal of Money, Credit and Banking 1: 15– 29. Google Scholar CrossRef Search ADS Tirole, J. 2006. The theory of corporate finance . Princeton, NJ: Princeton University Press. Townsend, R. 1979. Optimal contracts and competitive markets with costly state verification. Journal of Economic Theory 21: 265– 93. Google Scholar CrossRef Search ADS Whited, T., and Wu. G. 2006. Financial constraints risk. Review of Financial Studies 19: 531– 59. Google Scholar CrossRef Search ADS Weber, M. Forthcoming. Cash flow duration and the term structure of equity returns. Journal of Financial Economics . Footnotes 1 It is reasonable to assume that the choice of Arthur Andersen as an auditor is uncorrelated with the firm’s fundamental characteristics because the subpar accounting practices of Enron and Arthur Andersen did not surface until 2001. In fact, several studies find that Arthur Andersen clients were not more likely to be involved in financial restatements or earnings manipulation than other firms in the years preceding the Enron scandal (Dyck, Morse, and Zingales (2014)). 2 See Sellin (2001) for an earlier survey and the seminal work by Bernanke and Kuttner (2005), among others. 3 The Online Appendix shows this pattern for the sample of 1994–2008. 4 To address the immediate issue whether industry differences are enough to explain the differences across firms, the Online Appendix separately repeats Figure 1 for each industry and shows that there is still a large within-industry variation. 5 The recent paper by Chava and Hsu (2015) deviates from my findings about WW and HP indices. My sample goes further to mid-2008, when conventional monetary policy was still active, and therefore includes an additional recession where financial frictions were likely to play a significant role. Moreover, I use only scheduled FOMC meetings for reasons explained in Section 1.1. 6 Another reason for failure of the Kaplan-Zingales index might be that leverage enters this index with a positive sign, contradicting the fact that financially constrained firms typically have lower leverage, as discussedin Farre-Mensa and Ljungqvist (2015). In their words, “the KZ index is more of an outlier.” 7 In the words of Greenspan (Federal Open Market Committee (1995)), once the blackout period is over, “the thrust of the announced decision of the Committee then gets diluted.” 8 See, for example, Bernanke and Kuttner (2005) and Gorodnichenko and Weber (2016). 9 I discuss the leverage effect and other potential explanations in Section 3 in more detail. 10 A weaker, but still sufficient, version of this assumption is that the choice of auditor is independent of the characteristics that might affect how a firm’s stock price sensitivity to monetary policy shocks changes upon learning of the Enron scandal. 11 The behavior of leverage following the scandal is discussed in Section 3 in more detail. 12Ozdagli (2012) shows that book leverage can be calculated using the book-to-market equity ratio and market leverage. Therefore, I omit book leverage as an additional control to avoid multicollinearity. For size, I use book value of assets, rather than market value of equity for two reasons. First, the market equity size premium has declined significantly in the last three decades. Second, since Gertler and Gilchrist (1994), balance sheet (rather than market) size has been widely used in studies of the transmission of monetary policy. 13 Alternatively, one could use Fama-French factors based on these risk characteristics (size, market-to-book, and beta) to calculate returns not explained by these factors. However, directly controlling for firm characteristics subsumes the effect of the Fama and French (1992) risk factors. See, for example, Daniel and Titman (1997) and Ferson and Harvey (1999). 14 About 95% of Enron’s stock price decline between October 16 (Enron’s announcement of accounting losses) and November 8 (restatement) had already happened by November 6th. This suggests that the market had priced in the financial restatement before the FOMC announcement on November 6th. 15 Incidentally, other yields on the Treasury yield curve moved similarly as well. On May 15th and November 6th, 3-month yields moved $$-$$6 bp versus $$-$$5 bp, 1-year yields moved $$-$$3 bp versus $$-$$4 bp, and 2-year yields moved $$-$$3 bp versus $$-$$5 bp. 16 This 100 basis point stock price increase in response to a 10 basis point surprise reduction in the policy rate comes from a panel regression using all scheduled FOMC announcements in 2001. 17 When I include cash-flow duration and variation as control variables as well the coefficient remains practically the same, $$-$$0.64 versus $$-$$0.63, however the standard error increases to 0.38 ($$p<0.1$$), which is not suprising given that the inclusion of these variables causes loss of one third of the observations. In order to avoid this significant loss of data, I omit these variables in the remaining regressions. 18 In particular, about 95% of the firms retain their auditor from one year to the next, whereas firm characteristics, such as leverage, have an autocorrelation of about 50% to 60%. Market-based risk characteristics, such as book-to-market ratios, that are deemed to be more important determinants of stock returns have even lower autocorrelation. Therefore, while the auditor choice in 1995 is highly correlated with the auditor choice in 2000 (relevance condition), it would be much less correlated with any potential omitted firm characteristics that affect the reaction of stock prices to monetary policy shocks in 2000 (exclusion restriction). 19 Enron’s stock price was flat in the November 6 event window (9.69 versus 9.67), whereas on May 15 it went down by about 1%, from 57.45 to 56.99. Therefore, if there was any Enron-related news that would depress stock prices of Arthur Andersen clients in this window it is more likely to happen on May 15 rather than on November 6, which stacks the odds against my main result. The intraday approach also addresses any concern regarding pre-FOMC drift (Lucca and Moench (2015)). 20 To address any endogeneity concerns regarding ratings availability, I also use an instrumental variable approach employed in Faulkender and Petersen (2006) and Sufi (2007) where the instrumental variables are whether a firm is in the S&P 500, whether the firm is listed on the New York Stock Exchange, and whether the firm is in a three-digit SIC industry that includes other firms with credit ratings. The IV estimate for the difference between rated and unrated firms is $$-2.21$$ ($$p<0.01$$), and the Hausman test cannot reject the consistency of the OLS estimate ($$p=0.25$$). 21 Further details can be found here: http://www.moodysanalytics.com/~/media/Brochures/Credit-Research-Risk-Measurement/Quantative-Insight/CreditEdge/EDF-Expected-Default-Frequency-Overview.pdf. The monthly distress likelihood score of Campbell, Hilscher, and Szilagyi (2008) is also tried as an alternative measure. While the results are qualitatively similar, they have lower statistical significance, as would be expected from the increased measurement error due to matching daily data with monthly proxies. 22 A separate analysis using the Hodrick-Prescott filter leads to similar results. However, the moving-average of past values does not suffer from the look-ahead bias inherent in the Hodrick-Prescott filter. 23 For evidence that interest expense of financially constrained firms is also less responsive to monetary policy, see the Online Appendix. 24 To analyze the source of the change of leverage, the Online Appendix shows that the change in leverage comes from the decline in debt rather than changes in assets. Moreover, there is no evidence that the Arthur Andersen clients burn through cash relative to total assets faster than other firms, suggesting that it is not the use of internal funds that explains the leverage results. This might be because, among other things, that AA clients would like to hold on to cash as a precautionary motive due to their limited access to other financing. 25 The original BGG framework can be readily obtained by imposing $$E\left( w\right) =R^{K}$$, where $$R^{K}$$ is the return on capital and by replacing $$B=Z\left( QK-N\right) $$, where $$Z$$ is the contractual rate on capital. 26 In another famous model, Kiyotaki and Moore (1997), the firmborrows at the risk-free rate up to the collateral value of its capital. I focus on costly state verification because risky borrowing is a more realistic assumption formy sample. 27 In addition, E(w) might decrease for firms that become more difficult to audit, which would amplify the channel by reducing leverage. However, this may not be an important issue for this paper’s approach because Gleason, Jenkins, and Johnson (2008) find that contagion of financial restatements has economically small effects on near-term earnings forecasts of peer firms, and the earnings forecasts have very little explanatory power for the contagion in stock prices. 28 See, for example, Gorodnichenko and Weber (2016) and Ippolito, Ozdagli, and Perez (2017). © The Author 2017. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.
http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png
The Review of Financial Studies
Oxford University Press
http://www.deepdyve.com/lp/oxford-university-press/financial-frictions-and-the-stock-price-reaction-to-monetary-policy-06WJG0IXGr