Financial conditions and macroeconomic behavior.

What role do financial factors play in business cycles? Many business economists and policymakers, including the chairman of the Federal Reserve Board, have cited "balance sheet conditions" as a contributing factor in both the 1990-1 recession and the extended period of stagnant growth that surrounded it. Underlying this view is the belief that unusually weak balance sheets of nonfinancial companies, depository institutions, and households were constraining spending. When Alan Greenspan spoke repeatedly of a "50-mile-an-hour headwind" that was interfering with the recovery, he had in mind these kinds of financial factors.

Much of my research has been aimed at trying to understand the links between financial conditions and macroeconomic performance. Eight years ago, Ben S. Bernanke and I developed a simple business cycle framework in which endogenous fluctuations in borrowers' financial positions served to propagate the impact of exogenous disturbances to the economy.(1) We began with the idea that, for a significant class of borrowers, information and enforcement problems may drive the cost of uncollateralized external funds above the price of internal funds. Under these circumstances, borrowers' available supplies of collateral (broadly defined) and internal funds influence their spending and production decisions. In the aggregate, swings in borrowers' balance sheets over the cycle amplify fluctuations in spending and output. We referred to this propagation mechanism as a "financial accelerator."

One by-product of having financial conditions play a key role in the model is that the output dynamics are inherently nonlinear. Because the credit frictions bind across a wider cross section of borrowers in bad times (when balance sheets are relatively weaker on average) than in good times, contractions are typically sharper than expansions. Broadly speaking, this kind of nonlinearity appears consistent with the data.

In addition to rationalizing a new kind of business cycle propagation mechanism, the model provides a formal basis for Irving Fisher's "Debt Deflation" theory of the Great Depression. To explain the severity of the Depression, Fisher cited the sharp decline in prices--largely unanticipated, in his view--that greatly reduced the net worth of the borrowing class by raising the real value of outstanding debts. In our framework, a contraction in borrower net worth curtails borrowers' access to credit, inducing a persistent downturn.

Our early paper emphasized how financial positions might influence the behavior of nonfinancial companies. In principle, these conditions might influence other important classes of borrowers similarly, including (at least subsets of both) financial intermediaries and households. In our 1987 paper, we showed how a shortage of bank capital could induce a decline in lending by restricting banks' ability to attract uninsured deposits.[2] Many observers have maintained that this kind of phenomenon was particularly relevant during the recent capital crunch in banking.

Balance sheet effects on household spending are potentially quite important, in my view. Generally speaking, households face large spreads between borrowing and lending rates (after controlling for default probabilities), particularly for unsecured loans. In addition, some major household purchases, most importantly housing, are linked directly to the condition of household balance sheets by such features as downpayment requirements, up-front transactions costs, and minimum income standards. A recent example of a formal model in which household balance sheets influence housing demand is Edward C. Prescott.[3]

Our original theoretical framework was quite stark, designed mainly to make qualitative points. In subsequent work with Bernanke, with other coauthors, and by myself, I have enriched the institutional detail, in an attempt to move the theory closer to the data.[4] Some very recent work by others that makes progress along these lines includes Nobuhiri Kiyotaki and John Moore, Owen Lamont, and Jonas Fisher.[5]

In recent years I have turned my attention to the empirical side of the issue. Attempting to identify and quantify a financial accelerator mechanism in the data is a difficult task. Because existing datasets generally were not well suited for the problem, a large part of my effort has involved constructing an entirely new dataset.

One way to appreciate the obstacles present is to understand what will not work. Perhaps contrary to conventional wisdom, it is not possible to identify a financial accelerator effect in the data either by examining the forecasting power of credit aggregates or, more generally, by studying the lead/lag pattern between credit aggregates and output.

In the Bernanke-Gertler framework, for example, technology shocks are the only exogenous disturbances in the model. Credit conditions shape the dynamic response of output to these shocks, but they are not a primitive causal force. Thus, even though financial factors play an important role in output dynamics, a credit aggregate would have no marginal forecasting power for output, once technology shocks are included in the information set. A similar observation applies if monetary policy shocks also were a primitive driving force. Once indicators of monetary policy shocks are accounted for in the information set, we should not expect credit aggregates to have any marginal forecasting power.

Assessing the timing relationships between credit aggregates and output also is not helpful. The theory does not predict that credit aggregates should lead output over the cycle. Indeed, there is likely to be a countercyclical component to the demand for credit. Both households and firms may desire to borrow to smooth the impact of transitory variation in their incomes over the cycle. For example, as cash flows decline at the onset of a recession, firms may want to borrow to finance the buildup of unsold inventories and other fixed short-term obligations.[6] Firms thus may increase their borrowing temporarily in the early stages of the downturn, even if they have imperfect access to credit (The credit market frictions do not imply that firms are unable to borrow. Rather, they imply only that they will borrow less than they would otherwise, relative to a setting of perfect markets.) As a consequence, credit aggregates may lag rather than lead the cycle, even when financial conditions are shaping output dynamics. The Kiyotaki-Moore framework provides a nice formalization of this point.

I have approached the identification problem by exploiting a combination of both cross-sectional and temporal implications of the theory. The cross-sectional implications suggest sorting borrowers according to their relative access to credit, and then looking for differences in behavior implied by the theory, after controlling for nonfinancial sources of heterogeneity. Microdata studies of liquidity constraints commonly have employed this identification strategy, both for the case of households (for example, Stephen Zeldes)[7] and for the case of firms (for example, Stephen Fazzari, R. Glenn Hubbard, and Bruce Peterson).[8] The temporal implications suggest examining how the behavior of borrowers may vary over different phases of the business cycle. For many borrowers, credit market frictions are more likely to impinge on behavior around recessionary periods, when their balance sheets are weak, than around booms, when they are more likely to have adequate supplies of collateral assets and internal funds.

Hubbard and I used a panel dataset of individual manufacturing firms to identify a financial accelerator effect on investment.[9] We exploited both the cross-sectional and temporal implications that I have just described. We first grouped firms according to their relative access to credit, using dividend behavior as a criterion. We then estimated investment equations that allowed for differences in behavior across groups of firms and or across different phases of the business cycle. We found that, after allowing for non-financial influences, liquidity mattered more for the firms that were more likely to be constrained a priori. Further, the influence of liquidity on investment for these firms was stronger around recessionary periods than during booms, in keeping with the theory. Several recent studies have found similar evidence of an asymmetric impact of liquidity constraints over the cycle.[10]

While analysis of panel data provides a useful way to approach the identification problem, there are some limitations to existing studies. First, the samples are typically not representative of the relevant population of borrowers. This makes it difficult to draw inferences about the importance for aggregate activity. Second, the data typically are available only at the annual frequency. Therefore, dynamics at the business cycle frequency are hard to capture.

For these reasons, Simon Gilchrist and I decided to construct a panel dataset using information available from the Quarterly Financial Reports (QFR) published by the Census Bureau. The advantages of the QFR are that it has reasonably comprehensive cross-sectional information for several important sectors, including manufacturing, and that it is available at the quarterly frequency over a long period, from 1958:Q1 to the present. The main disadvantage of the QFR is that, until recently, it disaggregated only by size class. However, within the last year, we obtained the raw firm-level data that the QFR uses to construct the size class aggregates. Currently we are working with this data.

My initial investigation with Gilchrist involved comparing the cyclical behavior of the size class aggregates for manufacturing firms.[11] Our goal was to develop a set of basic facts to guide future research. We first reaggregated the firm size class variables into two categories: "small-firm" and "large-firm." For guidance, we used available information on financing patterns. Under our classification scheme, small firms rely heavily on intermediated credit. Further, they obtain about 80 percent of their short-term credit from banks and do not issue commercial paper. Large firms primarily obtain credit directly from the open market. And, they account for nearly all of the outstanding issues of commercial paper by manufacturers. Our small-firm category also squares reasonably well with the evidence from microdata studies: the largest firms in our small-firm category are similar in size to the median of the "liquidity-constrained" firms in the typical panel data study. Overall, by our criteria, small firms account for about 30 percent of total manufacturing sales.

We then traced the impact of a shift to tight monetary policy on the time-series behavior of small firms relative to large firms.[12] We examined sales, inventories, and short-term debt. Overall, small firms contract substantially relative to large firms. Ten quarters following a shift to tight money, small firms are typically down 17 or 18 percentage points relative to trend (measured by sales behavior), while large firms are down only about 6 or 7 percentage points.

The differences in inventory behavior and short-term borrowing are more striking. As their sales begin to decline, large finns appear to borrow to smooth production. Their inventories initially rise. So does their short-term borrowing. In contrast, after a brief period, small firms quickly shed inventories and contract their short-term borrowing. The difference in the percentage cumulative drop in inventories across the size classes is roughly 20 points, about twice the difference in the drop in sales. The difference in the cumulative drop in short-term debt is even greater: between 25 and 30 points. (Interfirm trade credit does not appear to adjust to offset the relative drop in short-term borrowing by small firms.[13] Trade credit to small manufacturing firms drops at about the same pace as short-term borrowing. Further, net trade credit [payables minus receivables] does not rise.)

To try to sort financial from possible nonfinancial explanations for our results, we performed two different kinds of exercises. First, we presented evidence of asymmetries in small-firm inventory/sales dynamics over the cycle. In booms, small firms appear to smooth production, much like large firms. It is mainly in recessions that small firms shed inventories quickly as sales drop. Large finns do not exhibit this kind of asymmetry. Second, we estimated some simple inventory equations that permitted technological coefficients to vary across the size classes. We found that balance sheet positions significantly influenced inventory investment for small firms, but not for large firms. In the estimation, we controlled for the possibility that movements in balance sheets may simply be signaling profits.

The most definitive way to pin down the influence of financial effects, of course, is to use microdata. As I mentioned earlier, we are now analyzing the QFR firm-level data. The first stage has involved mainly descriptive analysis, in order to understand the data. In a recent paper with Bernanke, we showed that the cyclical differences across the size classes that emerged in our earlier work largely remain intact, even after controlling for industry differences.[14] We also showed that similar cyclical differences emerge when we use a financial indicator to sort finns (specifically, a measure of bank dependency), rather than size.

While we have focused on inventory behavior, several recent studies have found that very similar differences in cyclical behavior across the size classes emerge for both investment and employment.[15]

The share-weighted differences between small and large finns in both sales and inventory fluctuations are significant relative to the manufacturing total. It is possible, however, that production could be shuffled from one class of firms to another without any impact on aggregate output. However, as with any "sectoral shocks" theory, aggregate effects will emerge if either: factors are not perfectly mobile in the short run; outputs are not perfectly substitutable; or workers and owners are not perfectly insured against the sectoral shocks. On the other hand, our numbers may understate the true aggregate impact if there are aggregate demand externalities (see, for example, Lamont) or factor-market linkages. This suggests that a better understanding of the connection between industrial structure and the business cycle is necessary.

To pin down the aggregate importance of the kinds of phenomena I have been describing, it is also crucial to examine data outside the manufacturing sector. While firms with imperfect access to credit (by my criteria) account for roughly 30 percent of manufacturing output and employment, they may account for somewhere between 40 to 50 percent of total output. And they are heavily concentrated in some important cyclical sectors, such as retail and wholesale trade and construction. For example, finns that do not have a rating to issue publicly traded debt account for roughly 80 percent of output, employment, and inventories in retail trade. This is especially significant, given the importance of retail inventories in the business cycle.

More generally, it is important not to fall into the trap of thinking about the manufacturing sector in isolation of other sectors. Firms in trade, construction, and services, for example, purchase goods from manufacturers. To the extent that financial conditions influence the spending of these firms, they may contribute to manufacturing fluctuations. In a similar vein, it is important not to ignore the household sector. Gathering evidence on the impact of financial positions on household spending, particularly spending on housing and other durable goods, is potentially a very important task.

1 B. S. Bernanke and M. Gertler, "Agency Costs, Collateral, and Business Fluctuations," American Economic Review 79, 1 (March 1989), pp. 14-31.

2 B. S. Bernanke and M. Gertler, "Banking and Macroeconomic Equilibrium," in New Approaches to Monetary Economics, W. Barnett and K. J. Singleton, eds. Cambridge, U.K.: Cambridge University Press, 1987.

3 E. C. Prescott, "Effects of Unanticipated Monetary Policies: An Unanticipated Finding," University of Minnesota, mimeo, 1993.

4 B. S. Bernanke and M. Gertler, "Financial Fragility and Economic Performance," Quarterly Journal of Economics (February 1990), pp. 87-114; M. Gertler, "Financial Capacity and Output Fluctuations in an Economy with Multiperiod Financial Relationships," Review of Economic Studies (July 1992), pp. 455-472; M. Gertler and R. G. Hubbard, "Corporate Financial Policy, Taxation, and Macroeconomic Risk," Rand Journal of Economics 24 (Summer 1993), pp. 286-303; and M. Gertler and K. Rogoff, "North-South Lending and Endogenous Domestic Capital Market Inefficiencies, "Journal of Monetary Economics 26,2 (September 1989).

5 N. Kiyotaki and J. Moore, "Credit Cycles," mimeo, 1993; O. Lamont, "Corporate Debt Overhang and Macroeconomic Vulnerability," MIT, mimeo, 1993; and J. Fisher, "Credit Market Imperfections and the Heterogeneous Response of Firms to Monetary Shocks," mimeo, 1994.

6 For evidence of a countercyclical component to short-term business borrowing, see: M. Gertler and S. Gilchrist, "The Role of Credit Market Imperfections in the Monetary Transmission Mechanism: Arguments and Evidence," Scandinavian Economics Journal (1993), pp. 43-64, and "The Cyclical Behavior of Short-Term Business Lending: Implications for Financial Propagation Mechanisms," European Economic Review (1993); and L. J. Christiano, M. Eichenbaum, and C. Evans, "What Happens After a Monetary Policy Shock? Evidence from Flow of Funds," mimeo.

7 S. Zeldes, "Consumption and Liquidity Constraints: An Empirical Investigation," Journal of Political Economy (1990), pp. 275-298.

8 E. C. Prescott, "Effects of Unanticipated Monetary Policies:..., op. cit., and S. Fazzari, R. G. Hubbard, and B. Peterson, "Financing Constraints and Corporate Investment," Brookings Papers on Economic Activity 1 (1988), pp. 141-195.

9 M. Gertler and R. G. Hubbard, "Financial Factors in Business Fluctuations," in Financial Market Volatility, Kansas City, Mo.: Federal Reserve Bank of Kansas City Annual Symposium, 1988, pp. 33-71.

10 A. K. Kashyap, O. Lamont, and J. C. Stein, "Credit Conditions and the Cyclical Behavior of Industries," Quarterly Journal of Economics, forthcoming, and S. Oliner and G. Rudebusch, "Is There a Broad Credit Channel?" Federal Reserve Board of Governors, mimeo.

11 M. Gertler and S. Gilchrist, "Monetary Policy, Business Cycles, and the Behavior of Small Manufacturing Firms," Quarterly Journal of Economics (May 1994).

12 Credit conditions may influence the transmission of monetary policy via the financial accelerator. For a discussion, see M. Gertler and S. Gilchrist, "Monetary Policy, Business Cycles..., op. cit. A complementary way is via the "lending channel." See A. K. Kashyap and J. C. Stein, "Monetary Policy and Bank Lending," in Monetary Policy, N. G. Mankiw, ed. Chicago: University of Chicago Press, 1994.

13 See M. Gertler and S. Gilchrist, "The Role of Credit Market Imperfections..., op. cit.

14 B. S. Bernanke, M. Gertler, and S. Gilchrist, "The Financial Accelerator and the Flight to Quality," mimeo, December 1993.

15 S. Oliner and G. Rudebusch, "Is There a Broad Credit Channel? . . . op. cit., and S. Sharpe, "Financial Market Imperfections, Firm Leverage, and the Cyclicality of Employment," American Economic Review, forthcoming.

Mark Gertler has been a research associate of the NBER since 1990. He is also a professor of economics at New York University and, beginning in September, will serve as an academic consultant at the Federal Reserve Bank of New York.

Gertler received a B.A. from the University of Wisconsin and a Ph.D. from Stanford University. He has taught at Cornell University, the University of Wisconsin, Stanford, and Columbia. His work on macroeconomics, monetary policy, and financial markets has been published in numerous professional journals and in the NBER Working Paper series. He is also an associate editor of the Journal of Money, Credit and Banking.

Gertler and his wife, Cara Lown, live in Manhattan. She is an economist at the Federal Reserve Bank of New York. Lately, Gertler has been spending his free time watching the Knicks and preparing for fatherhood. He and his wife are expecting a baby girl to arrive at about the time of publication of this issue of the NBER Reporter.