U.S. Monetary Shocks And Global Stock Prices

5We measure an industry’s financial dependence also using U.S. data, following theinfluential work by Rajan and Zingales (1998), to gauge the intrinsic demand for external financein the absence of financial constraints. This approach relies on the assumption that large U.S.listed firms face minimal financial constraints given the depth of U.S. financial markets and thatthe ranking of financial dependence across sectors in the U.S. is preserved in other countries.7We find strong evidence of a negative response of stock prices to U.S. monetary shocks,with U.S. monetary policy loosening (tightening) being associated with an increase (decrease) instock prices in other countries, consistent with earlier work based on U.S. stock prices (e.g.,Thorbecke, 1997, and Bernanke and Kuttner, 2005). Moreover, this impact is particularlypronounced for firms with a relatively high intrinsic dependence on external finance. Forexample, an unexpected policy rate decrease of 5 basis points (equal to its interquartile range ) isassociated with a stock price response that is 6 basis points greater for firms whose financialdependence is at the 75th percentile (the Construction machinery industry) relative to firmswhose financial dependence is at the 25th percentile (the Beverages industry). This is asignificant effect compared to the average stock market return around FOMC dates of 19 basispoints.To further distinguish financial from other explanations, we consider asymmetries overthe business cycle. Financial constraints are likely to be binding for more firms duringrecessions. We therefore expect our effect to be stronger during economic downturns. We indeedfind that the effect is stronger during U.S. recession periods.Finally, we examine if the impact of U.S. monetary shocks on stock prices varies acrosscountries, based on country features that capture differences in access to financial markets, suchas financial integration and development. We find that the impact of monetary shocks on stockprices is more pronounced in countries that are more financially integrated with the rest of theworld, where we measure financial integration following Lane and Milesi-Ferretti (2007) as acountry’s foreign assets and liabilities over GDP.Taken together, these results suggest that financial frictions play an important role in thetransmission of monetary policy, and that U.S. monetary policy influences global assetallocation.Empirical research on the link between monetary policy shocks and stock prices hasgenerally not considered the role of financial constraints. Bernanke and Kuttner (2005) andBoudoukh et al. (1994) analyze the differential impact of monetary policy shocks on stock pricesacross broad classes of industries but do not explicitly consider the role of financial constraints,while Kashyap et al. (1994) do consider financial constraints to show that firm inventoryinvestment by liquidity constrained firms is significantly adversely affected during periods of7Following Rajan and Zingales (1998), who use this approach to study the impact of financial development oneconomic growth, this approach has been applied, among others, to study the role of business cycles (Braun andLarrain, 2005), demand for working capital (Raddatz, 2006), and financial crises (Kroszner et al., 2007) ininfluencing the link between finance and growth.

6tight monetary policy but they do not analyze its impact on stock prices. Similarly, Gertler andGilchrist (1994) show that the investment of small firms, their proxy for the importance offinancial frictions, responds more strongly to monetary policy than that of large firms. Finally,research on stock prices and financial constraints has generally not considered the role ofmonetary policy (e.g., Baker et al., 2003). An exception is Lamont et al. (2001) who find littlerole for monetary policy but they use traditional measures of monetary policy that do notdisentangle monetary shocks from market expectations.The paper proceeds as follows. Section 2 presents our empirical strategy, construction ofkey variables, and sources of data. Section 3 discusses the main empirical results and a slew ofrobustness checks and extensions. Section 4 offers concluding remarks.II. METHODOLOGY AND DATAA. MethodologyOur basic empirical strategy is to use event study analysis to test whether an exogenousmonetary shock in the U.S. has an impact on the stock return of firms in other countries, andwhether this effect is more pronounced for firms that are more financially dependent. We use theapproach in Bernanke and Kuttner (2005) to identify unexpected policy rate changes, and buildon their work by extending the analysis to other countries and by considering asymmetricresponses across firms depending on their degree of financial dependence. This allows us to dealmore effectively with concerns about endogeneity and simultaneity, and discern more preciselyone of the channels through which monetary policy affects stock prices.Our analysis starts by confirming the common finding in the literature that stock returnsare negatively associated with innovations in monetary policy. We do this by showing that stockprices respond negatively to unanticipated changes in the U.S. Federal funds rate followingmeetings of the US Federal Open Market Committee (FOMC). To be precise, we estimate thefollowing equation:Stock Return i , j , k ,t Monetary Shock t Controli ,t i , j ,k ,t(1)where i stands for company, j for country, k for sector, and t for time. Note that this is a panelregression. We start by assuming the same β for all sectors and countries in order to estimate anaverage effect, but will subsequently allow for variations across sectors, countries, and time. Weinclude firm specific fixed effects to control for unobserved firm specific factors, and clusterstandard errors at the FOMC date level, to adjust standard errors for cross-sectional correlationover time.Asset pricing models offer guidance for the inclusion of control variables. In the basespecification, we include the three Fama and French (1992) factors: firm size (log assets), theratio of the market value to book value, and the beta coefficient (i.e., the correlation between thefirm’s stock return and the market return) times the stock market return. These control variables

7are lagged by one-year, except the stock market return which we include contemporaneously, toalleviate concerns about endogeneity. We follow Whited and Wu (2006) and incorporate thesethree factors by entering the relevant firm characteristics directly into our regressions rather thanby first estimating a factor model. For our purposes, these two alternative ways of incorporatingthe three factors are equivalent. Entering firm characteristics directly in our regressions is easierto implement, though the interpretation of the coefficients on these factors is lessstraightforward.To investigate how an industry’s financial dependence affects the impact of the U.S.monetary policy shock, we now consider the interaction between the monetary policy shock andan industry’s dependence on external finance. In other words,βk β 1 β2 Financial Dependencek(2)where Financial Dependencek measures the external financing needs for capital expenditure forfirms in a given industry following Rajan and Zingales (1998). The slope coefficient, β2, thencaptures the extent to which the effect of the monetary policy shock depends on an industry’sdependence on external financing. We include firm and time specific fixed effects to control forunobserved firm and time specific factors, and cluster standard errors at the time level, to adjuststandard errors for cross-sectional correlation at different dates.To further distinguish financial from other explanations, we consider asymmetries overthe business cycle. Financial constraints are likely to be binding for more firms duringrecessions. We therefore expect out effect to be stronger during economic downturns. We testthis prediction by including an interaction between the Monetary Shock * Financial Dependencevariable and a Recession variable that denotes whether or not the FOMC meeting date occursduring a recession.In other words, we extend equation (2) as follows:βkt β 1 β2 Financial Dependencek β3 Financial Dependencek * Recessiont(3)where Financial Dependencek measures the external financing needs for capital expenditure forfirms in a given industry following Rajan and Zingales (1998) and Recessiont indicates FOMCmeeting dates that fall during recession periods. The slope coefficient β2 then captures the extentto which the effect of the monetary policy shock depends on an industry’s dependence onexternal financing during non-recession periods, and the slope coefficient β3 then captures theextent to which this effect differs during recession periods. We expect negative signs for bothcoefficients. Note that the inclusion of the Recession variable makes the coefficient on themonetary shock interaction variable time dependent.Finally, we examine if the impact of U.S. monetary shocks on stock prices varies acrosscountries by including additional interaction terms between country characteristics (such asfinancial integration and financial development) and the monetary shock variable. In otherwords, we extend equation (2) as follows:

8βkjt β 1 β2 Financial Dependencek β4 Country Traitjt(4)where the slope coefficient β4 then captures the extent to which the effect of the monetary policyshock depends on a particular country trait.B. Data and Variable DefinitionsStock pricesTo construct our dependent variable, we collect data on stock prices of 20,121 firms in 44countries over the period 1990 to 2008. Appendix Table 1 shows the complete list of countries.Stock price data are retrieved from Datastream, and are adjusted for dividends and capital actionssuch as stock splits and reverse splits. We consider two-day stock price responses to monetarypolicy shocks following FOMC meetings. Specifically, we compute the stock return as the logdifference in the closing price of the stock over the period t-1 and t 1, where t is the day of theFOMC meeting. The reason for using a two-day window rather than a one-day window of stockreturns is due to time zone differences between stock markets in the U.S. and other countries. Toreduce the impact of extreme values, we drop two-day stock returns with a value of above 50%or below -50%, which covers 0.1% of the sample. As a robustness check, we also winsorize thesample at its top and bottom 1% level, and the key results carry through. Our sample includes ofa total of 140 FOMC meetings, for a total of 925,306 firm-date observations.Monetary Policy ShockOur measure of monetary shocks at U.S. FOMC meetings follows the approach inKuttner (2001) and Bernanke and Kuttner (2005).8 They propose to use the change in the price offederal funds futures contracts relative to the day prior to the policy action as a measure ofunexpected policy rate changes.9 For a FOMC meeting on day d of month m, the monetary shockis the change in the rate implied by the current-month futures contract. However, because thecontract’s settlement price is based on the monthly average federal funds rate, the change in theimplied futures rate then should be scaled up by a factor related to the number of days in themonth affected by the change, or:8Alternative measures of monetary policy include those developed by Bernanke and Mihov (1998) and Romer andRomer (2004), among others. The measure of monetary policy by Bernanke and Mihov is computed using VARmodels on monthly data and is therefore not applicable to our event study setting for which daily data are needed.The Romer and Romer measure has the right frequency but uses information from the Fed’s greenbooks that includeFed staff economic forecasts that are not made publicly available to the market until five years after the FOMCmeeting and may therefore not be fully incorporated in stock prices.9This approach assumes that risk premia that could be embedded in prices on federal funds futures do not changesystematically within a one day period (Piazzesi and Swanson, 2008).

9Shockd D f d f d 1 D dwhere Shock is the unexpected target rate change, f d is the current-month futures rate at day d,and D is the number of days in the month. The expected rate change then is the actual changeminus the shock.We extend the data coverage of monetary shocks from year 2002 to year 2008. Moreover,while Bernanke and Kuttner (2005) exclude FOMC dates when there is no policy rate change,we include these dates as well as control groups. Appendix Table 2 lists the exact dates ofFOMC meetings, the actual changes of federal funds rate, and the unexpected component of thechange.We drop U.S. firms from the sample, since we use U.S. monetary shocks as our source ofexogenous variation in monetary policy. This lends additional credibility to using our measure ofmonetary shocks as an exogenous variable of monetary policy given that U.S. monetary policy isunlikely to respond in a systematic way to idiosyncratic economic factors in other countries,though it may respond to economic factors in the U.S.Financial dependence indexAs measure of an industry’s intrinsic dependence on external finance, we use thefinancial dependence measure proposed by Rajan and Zingales (1998). They compute anindustry’s dependence on external finance as:Financial dependence Capital expenditures - Cash flow,Capital expenditureswhere cash flow cash flow from operations decreases in inventories decreases inreceivables increases in payables. The index is computed using data on publicly listed U.S.firms, which are judged to be least likely to suffer from financing constraints relative to generallysmaller, non-listed U.S. firms and firms in other countries. Conceptually, the Rajan and Zingalesindex aims to identify sectors that are naturally more dependent on external financing for theirbusiness operation. While the original Rajan and Zingales (1998) paper covers only 40 (mainlySIC 2-digit) sectors, we recompute their measure using data for the period 1990-2006 to expandthe coverage to around 150 SIC 3-digit sectors. We drop firms active in the utilities industry(SIC 4), wholesale and retail industry (SIC 5), financial industry (SIC 6), and publicadministration (SIC 9) because these firms are subject to strict regulation or because theirfinancing needs are not comparable with those of other industries.To calculate the demand for external financing of U.S. firms, we take the following steps.We first sort every firm in the Compustat USA files based on their 3-digit SIC sectoralclassification and then calculate the ratio of dependence on external finance for each firm byaggregating cash flows and expenditures as in Rajan and Zingales over the period 1990-2006.

10We then calculate the financial dependence index as the sector-level median value of these firmratios for each SIC 3-digit sector that contains at least 5 firm observations.Recession datesWe use NBER recession dates, available on a quarterly frequency, to construct a dummyvariable Recession that indicates whether or not the data on which an FOMC meeting takes placefalls during a recession quarter.Control variablesThe three Fama and French (1992) factors that we include as control variables arecomputed using data from Worldscope and Datastream. We compute firm size as the log of totalassets, and the market-to-book ratio as the ratio of the market value of equity to the book valueof equity. The firm-level beta coefficient (that we interact with the stock market return) iscalculated as the correlation between the firm’s weekly stock return and the weekly country-levelreturn on the local stock market index. We use the domestic beta rather than a beta based on aworld factor model because Griffin (2002) finds that domestic factor models perform better inexplaining time-series variations in returns and have lower pricing errors than the world factormodel.C. Descriptive StatisticsTable 1 reports summary statistics of the key dependent and explanatory variables. Theactual change in the federal funds rate announced following FOMC meetings ranges from a ratecut of 100 bps to a rate increase of 75 bps. Unexpected rate shocks vary from -43 bps to 24 bps,indicating that rate surprises on average were on the downside.Financial dependence ranges from a low of -2.4 for the Manifold Business Formsindustry (an industry that has been in decline globally for over a decade and hence seencorrespondingly low investment) to a high of 1.4 for the Photographic Equipment and Suppliesindustry (an industry that has gone digital and hence seen large capital investment). Financialopenness, measured as a country’s foreign assets and liabilities over GDP, ranges from a low of0.3 percent for the Republic of Korea in year 1991 to a high of 23.9 percent for Hong Kong in2007.III. EMPIRICAL RESULTSA. Basic SpecificationThe baseline results are presented in Table 2. There we include firm fixed effectsthroughout all regression specifications, and cluster standard errors at the level of FOMCmeeting dates. In Column 1, we first examine the impact of actual federal funds rate changes onstock returns, to allow comparison with existing results in the literature. We obtain a negativecoefficient on the actual federal funds rate variable but this coefficient is not significantlydifferent from zero.

11In Column 2, we further decompose the change of the federal funds rate into its expectedand unexpected components, following the method proposed by Bernanke and Kuttner (2005).We find that the unexpected component has a significantly negative coefficient, suggesting thatunexpected monetary tightening reduces stock prices. Based on the estimated coefficient of 0.04,a 25-base point increase of U.S. rate would reduce global stock prices by about 1%. This is not atrivial number. Moreov

In this paper, we test this prediction using data on 20,121 firms in 44 countries by . stock prices in other countries, consistent with earlier work based on U.S. stock prices (e.g., Thorbecke, 1997, and Bernanke and Kuttner, 2005). Moreover, this impact is particularly