FINANCING CONSTRAINTS AND INVENTORY INVESTMENT: A COMPARATIVE STUDY WITH HIGH-FREQUENCY PANEL DATA Robert E. Carpenter, Steven M. Fazzari, and Bruce C. Petersen* Abstract —This study provides new evidence of the importance of ﬁnancing constraints for explaining the dramatic cycles in inventory investment. We compare the empirical performance of different ﬁnancial variables (coverage ratio, cash stocks, and cash ﬂow) used in previous research to test for the presence of ﬁnancing constraints. The comparison is undertaken in a common framework with an identical sample and high-frequency (quarterly) ﬁrm panel data. Cash ﬂow is much more successful than cash stocks or coverage in explaining the facts about inventory investment across ﬁrm size, different inventory cycles, and different manufacturing sectors. I. Introduction I N THE LAST decade there has been a dramatic revival of research on ﬁnancing constraints and ﬁrm behavior. The new literature covers a broad range of issues, including inventory investment, R&D, physical investment, pricing under imperfect information, business formation and sur- vival, tax policy, the business cycle, and the transmission of monetary policy. 1 This paper makes two principal contribu- tions to this literature. First, we use high-frequency (quar- terly) ﬁrm panel data to provide new evidence supporting the importance of ﬁnancing constraints for inventory invest- ment. Second, we provide the ﬁrst comparative study of different ﬁnancial variables used to test for the presence of ﬁnancing constraints. Contributing to the explanation of inventory cycles is a challenge for the ﬁnancing constraint literature. Inventory investment ﬂuctuations account for a surprisingly large fraction of the aggregate cycle. For example, Blinder and Maccini (1991) report that declines in inventory investment averaged 87% of the drop in aggregate U.S. output during postwar recessions. Other researchers in the inventory literature (e.g., Lovell (1994, p. 34)) point to the potential inﬂuence of ﬁnancing constraints on inventory investment as a major unanswered question. In addition to its aggregate volatility, there are important sectoral differences in the cyclical pattern of inventory investment. Stanback (1962, p. 23) and Zarnowitz (1985, p. 527) identify much larger cyclical movements in durable inventory investment compared with nondurables. Figure 1 extends Stanback’s evidence to the present, showing quar- terly growth rates of real inventory stocks for durable and nondurable manufacturing. 2 The durable series clearly dis- plays greater cyclical volatility. We explore the ability of the ﬁnancial variables to help explain this sectoral heterogene- ity, a new test of the ability of ﬁnancing constraints to explain inventory behavior. The studies that motivate our paper are Gertler and Gilchrist (1994), hereafter GG, Kashyap et al. (1994), hereafter KLS, and Carpenter et al. (1994), hereafter CFP. 3 Each study employs different ﬁnancial variables and econo- metric approaches and each emphasizes a different channel through which ﬁnancing constraints operate. GG employ the coverage ratio, a measure of ﬁrms’ ability to meet interest payments. Bernanke and Gertler (1995, p. 38) state that the coverage ratio is a good measure of the strength of a ﬁrm’s balance sheet and cite the GG ﬁndings as important evidence supporting the presence of a ‘‘balance sheet channel’’in the transmission mechanism for monetary policy. KLS include the stock of cash in an inventory regression. Their study is the ﬁrst to provide micro evidence in support of a ‘‘bank lending channel’’ in the transmission mechanism. 4 In CFP we focus on the impact of cash ﬂow for inventory invest- ment. We emphasize that large ﬂuctuations in cash ﬂow over the business cycle may cause ﬁrms to make large adjust- ments to inventories (a liquid, readily reversible investment with low adjustment costs) to partially offset shocks to cash ﬂow, the primary source of ﬁnance for most ﬁrms. We compare the performance of these three ﬁnancial variables in a common econometric framework with an identical sample of quarterly ﬁrm panel data. This kind of data provides several advantages. Quarterly data are desir- able to study a high-frequency phenomenon like inventory investment, yet few studies of ﬁnancing constraints have utilized it. 5 Our sample improves upon the data used by GG (aggregate times series, disaggregated into two ﬁrm-size classes) and KLS (individual annual cross sections). With panel data we control for individual ﬁrm effects that are likely to bias results from cross-sectional regressions. In addition, the large number of ﬁrms combined with the high-frequency data allows us to examine brief periods (two to four years), permitting multiple comparisons of the performance of the ﬁnancial variables over different cyclical Received for publication September 3, 1996. Revision accepted for publication November 26, 1997. * Emory University; Washington University and the Jerome Levy Economics Institute; and Washington University, respectively. We thank the Levy Institute for ﬁnancial support and Ben Herzon for excellent research assistance, and we acknowledge the helpful comments of Lee Benham, Robert Chirinko, Mark Gertler, Simon Gilchrist, Charles Himmelberg, Glenn Hubbard, John Keating, Michael Lovell, Louis Maccini, Alistair Milne, Donald Morgan, Dorothy Petersen, Steven Sharpe, and Victor Zarnowitz. We also received helpful comments from seminar participants at Dalhousie University and the National Autonomous University of Mexico and the 1996 ISIR-ASSA, Midwestern Economics, Eastern Finance, and Western Finance Association meetings. In addition we thank the editor and two anonymous referees for valuable comments. 1 See Hubbard (1998) for a review of the literature. 2 Both series in the ﬁgure are three-quarter moving averages to make the cyclical patterns more obvious. 3 Other empirical research on inventory, investment, and ﬁnancing constraints includes Calomiris et al. (1995) and Milne (1995). 4 This study has been inﬂuential in establishing a link between monetary policy and inventory investment that does not rely on an interest rate effect. See the discussion in the NBER Reporter (Fall 1995). 5 The evidence in Carpenter and Levy (1998) supports this point. They estimate that over three-quarters of the variation in monthly, industry-level inventory investment is contained in the high-frequency bands of the spectrum and cannot be detected with annual data. [ 513 ]  1998 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/003465398557799 by guest on 14 October 2021