Introduction Directory UMM :Data Elmu:jurnal:J-a:Journal of Economics and Business:Vol51.Issue3.May1999:

Nonmonetary Effects of the Financial Crisis in the Great Depression Ali Anari and James Kolari The credit hypothesis maintains that nonmonetary factors worsen declines in output during severe economic contractions, which has been a prominent rationale for stringent bank regulation. We apply recent advances in time series analysis to re-examine the role of U.S. bank failures in the Great Depression. In brief, month-by-month decompositions of output, prices, money supply, the liabilities of failed firms, and the deposits of failed banks indicate that bank failures did not initiate the fall in output and prices. However, chronic bank failure over at least a two-year period, in combination with a surge in failures in the latter part of this period of banking industry distress, had large negative effects on all of the variables under study. We conclude that chronic bank failures coupled with subsequent threshold failure effects can have deep and pervasive influences on the economy, which justify government intervention at such times. © 1999 Elsevier Sci- ence Inc. Keywords: Bank failure; Economic output; Great Depression JEL classification: E44, G21, N22

I. Introduction

Bernanke 1983 first proposed the credit hypothesis to help explain the collapse in U.S. economic productivity during the Great Depression. In brief, this hypothesis argues that bank failures have nonmonetary effects on the economy which disrupt credit flows in the financial intermediation process and, thereby, exacerbate declines in economic activity. An extreme case is the possibility of a banking crisis in which serious damage to the financial system is possible, i.e., loss of public confidence and resultant runs on insolvent banks that spread to solvent banks, which is known as the contagion effect [see Kaufman Center for Business and Economic Research, Texas A M University, College Station, Texas; Finance Department, Texas A M University, College Station, Texas Address correspondence to: Dr. James W. Kolari, Finance Dept., Texas AM University, College Station, TX 77843. Journal of Economics and Business 1999; 51:215–235 0148-6195 99 –see front matter © 1999 Elsevier Science Inc., New York, New York PII S0148-61959900005-3 1994]. To protect themselves, solvent banks not only shift funds to liquid assets but curtail lending, due to rising firm bankruptcies, which causes a lemons problem in the credit market [see Diamond and Dybvig 1983]. In this sense, bank failures are perceived as more damaging to the economy than the failure of other firms, and in the words of Kaufman 1996, p. 25, “ . . . almost all countries have imposed special prudential regu- lations on banks to prevent or mitigate such adverse effects.” Indeed, the Great Depression promulgated a panoply of U.S. banking regulations under the 1933 and 1935 Banking Acts which covered a broad array of issues, including pricing, products, and geographic expansion. Whereas the Great Depression witnessed the failure of some 9,000 banks between 1929 and 1933, and the loss of about 25 of bank system deposits [see Bernanke 1983 and Wheelock 1995], the post-Depression regulatory blanket preserved a more or less steady state of about 15,000 banks until 1980. At that time, increasing interest rate volatility, financial innovation, and new competition from nonfinancial firms began to motivate deregulation of Depression-era banking laws. Today, bank deposit and loan interest rates are totally deregulated, many restrictions on securities and insurance products are relaxed, and interstate banking is permitted. In turn, the number of bank failures per year has noticeably increased at times. For example, in the years 1987 to 1989, the number of failed banks plus failing banks involved in regulator- assisted mergers exceeded 200 per year [see Boyd and Graham 1996]. Recent work by Mason 1998 compared the direct and indirect costs of large or national bank failures in the Great Depression with the bank failure episode in the late 1980s and early 1990s in the United States. He calculated that FDIC disbursements to closed banks jumped in recent years, e.g., between 1988 and 1992, disbursements exceeded 10 billion per year and peaked in 1991 at about 20 billion. Mason 1998 found similarities in these two periods of banking distress, in terms of substantially lower recovery rates from closed banks, and concluded that potential macroeconomic losses due to asset market overhang andor inefficient management of failed bank assets can be significant. An important empirical issue in this respect is the sensitivity of real economic activity to bank failures. Do bank failures have immediate short-run negative effects on output, or are they associated with long-run effects? How important are the effects of bank failures on output relative to other factors, such as prices and money supply? Answers to these questions are relevant to the continued deregulation of the banking industry to the extent that there is compelling empirical evidence with respect to the credit hypothesis. With this motivation in mind, the present paper seeks to contribute further empirical evidence on the credit hypothesis by revisiting the Bernanke 1983 study. He and others [e.g., see Calomiris et al. 1986; Gertler 1988; Haubrich 1990] employed a single- equation time series regression approach to examine short-run interactions between key economic and financial variables. Unlike those studies, we have applied the more sophisticated econometric methods of vector autoregression VAR and cointegration analysis [see Johansen 1988, 1991]. These methods are superior to single-equation time series models because they fully incorporate the dynamic relationships among all the variables under investigation. Rather than basing our inferences on a few coefficients in a single regression equation viz., about 12 coefficients that reflects the relationships of the variables throughout the entire period of observation, we employed five equations which each contain five variables with 12 monthly lags. The historical decompositions of 1 For recent examples of studies employing vector autoregression and vector error correction, see King et al. 1991 and Kugler and Lenz 1993. 216 A. Anari and J. Kolari time series variables provide valuable information about the time path of the impact of each variable on the rest of the variables under investigation. Information about the month-by-month impact of each variable, with respect to the other variables in the model, are particularly useful in sorting out their relative contribution in explaining the declines in output, prices, and money supply. Based on data used in the Bernanke 1983 study, and historical decompositions of the time series variables in the VAR model, our results indicate that bank failures did not play an important role in explaining the drop in output until the last few months of the Great Depression episode. In this regard, persistent bank failures over at least a two-year period which escalated over time did contribute to output declines. Importantly, the VAR model results reveal that chronic and increasing bank failures had severe negative impacts on not only output but prices, money supply, and business failures. The policy implication of these findings to banking regulation is that bank failures by themselves do not necessarily have deleterious effects on the economy; however, chronic and severe bank failures seriously disrupt the financial intermediation process and potentially have pervasive adverse effects on the economy. Corroborating evidence for this inference is the finding by Mason 1998 that failed bank assets are liquidated more slowly and with lower recovery rates during periods of financial crisis. Consequently, at some point, government intervention is justified to stabilize the banking system. Although a direct link to current banking issues is not possible, these findings tend to suggest that, as long as there is a lender of last resort i.e., the central bank to overcome the problem of chronic and deepening failures, isolated failures of individual institutions in the context of today’s more competitive banking environment do not appear to be a major problem with respect to economic productivity. The next section overviews related empirical literature. Section III describes the Johansen 1988, 1991 cointegration procedure and the data. Section IV reports the empirical results, and the last section contains conclusions and implications.

II. Related Literature