The Asymmetric Effects of Financial FrictionsGuillermo L. Ordo˜nez⇤August 2012AbstractEconomic variables are known to move asymmetrically over the business cycle: quicklyand sharply during crises, but slowly and gradually during recoveries. Not known is thefact that this asymmetry is stronger in countries with less-developed financial systems. Thisnew fact is documented using cross-country data on loan interest rates, investment, andoutput. The fact is then explained using a learning model with endogenous flows of infor-mation about economic conditions. Asymmetry is shown to be stronger in less-developedcountries because these countries have greater financial frictions, which are captured inthe model by higher monitoring and bankruptcy costs. These greater frictions magnify thecrisis reactions of lending rates and economic activity to shocks and then delay their recov-ery by restricting the generation of information after the crisis. Empirical evidence and aquantitative exploration of the model show that this explanation is consistent with the data.⇤University of Pennsylvania, Department of Economics. E-mail: [email protected]. I especially thankAndy Atkeson, David K. Levine and Rob Shimer for excellent suggestions to improve the paper, Tommaso Porziofor outstanding research assistance, and Juan Manuel Licari and Nick Bloom for sharing data. For their comments,I also thank Ariel Burstein, V.V. Chari, Steve Durlauf, Christian Hellwig, Narayana Kocherlakota, David Lagakos,Hanno Lustig, Cesar Serra, Stijn van Nieuwerburgh, Laura Veldkamp, Mike Waugh, and seminar participantsat UCLA, the Minneapolis Fed, Minnesota, Wisconsin-Madison, Penn State, Yale, the 2006 Econometric SocietySummer Meeting (Minnesota), the 2006 SED Annual Meeting (Vancouver), the 2008 AAEP Meetings (Cordoba,Argentina), the 2008 LACEA-LAMES Meetings (Rio de Janeiro, Brazil), and the 2009 Conference on ”FinancialFrictions and Segmented Markets” at the UCSB. Finally, I appreciate the hospitality of the Federal Reserve Bank ofMinneapolis and the excellent editorial assistance of Kathy Rolfe and Joan Gieseke. The usual waiver of liabilityapplies.1 IntroductionEconomic variables are known to move asymmetrically during the bad and good phases of abusiness cycle: quickly and sharply during recessions, but slowly and gradually during recov-eries. Interest rates on loans, for example, tend to rise quickly during a crisis, but fall slowlyand gradually during a recovery; investment and output tend to move in the opposite direc-tions, but with the same asymmetry – falling sharply during a crisis and recovering slowly.This asymmetry has been observed worldwide in most business cycles. In Mexico’s 1994-95crisis, for example, real lending rates rose 70 percentage points in just four months and invest-ment and output per capita dropped 35% and 17%, respectively, in three quarters. Recovery ofthese variables took much longer: lending rates did not reach pre-crisis levels for 30 months;investment, for two years; and output, for almost three years.1Understanding the sources of this asymmetry would seem critical to minimizing lengthy pro-cesses of financial distress and the inefficient resource allocation inherent to lengthly recover-ies.2Not surprisingly, then, many studies have offered explanations for this asymmetry.3No one, however, has yet systematically examined how this business cycle asymmetry differsacross countries. I do that here using standard datasets and discover a new fact: the asymmet-ric movements of lending rates, investment and output are stronger in less-developed coun-tries, those with weaker financial systems. I also propose an explanation for this new fact: Thestronger asymmetry in less-developed countries stems from their greater financial frictions,which restrict the flow of information in an economy, delaying recoveries.I introduce a form of these frictions into a learning model with endogenous flows of informa-tion. Commonly in such models, the degree of precision of observed information depends onthe level of economic activity, which varies in good and bad times, thus generating asymmet-ric lending rates (Veldkamp (2005)) and asymmetric economic activity (Van Nieuwerburgh and1There are many other examples. In Brazil, just in October 1997, real lending rates rose from 71% to 98%, taking10 months to return to pre-crisis levels. The first quarter of 1998, investment and GDP per capita declined 9% and8% respectively, taking more than 2 years to return to pre-crisis levels. In Indonesia, the 8 months following theAsian crisis witnessed a rise in real lending rates from 18% to 35%, taking 24 months to recover. In that case, bothinvestment and GDP per capita declined almost 50% in 1998, recovering in more than 8 years. In Russia, duringApril and May of 1998, real lending rates rose from 30% to 150%, taking 27 months to recover, while GDP per capitadeclined 12% in one quarter, recovering in more than 2 years.2The negative impact of lengthy recoveries in terms of misallocations has been discussed by Bergoeing, Loayza,and Repetto (2004), while the sizable macroeconomic effects of those misallocations have been estimated by Hsiehand Klenow (2009) and calibrated by Restuccia and Rogerson (2008) and Buera, Kaboski, and Shin (2011).3Veldkamp (2005) studies asymmetries of lending rates while Van Nieuwerburgh and Veldkamp (2006) andJovanovic (2006) focus on asymmetries of real activity. The bulk of the literature, however, has focused on expla-nations of asymmetries in stock markets. Banerjee (1992), Welch (1992) and Banerjee and Newman (1993) explaincrashes as the result of herd behavior and information cascades. Jacklin, Kleidon, and Pfleiderer (1992) use a port-folio insurance model of stock market crashes. Allen, Morris, and Shin (2006) study information-based bubbles.Finally, Zeira (1994, 1999) proposes an informational overshooting to explain booms and crashes in stock prices.1Veldkamp (2006)). These models, however, assume frictionless environments in which lendersand borrowers have symmetric information about both aggregate and idiosyncratic shocks. Inmy model, I assume asymmetric information about idiosyncratic shocks; they are observableto borrowers for free and observable to lenders only at a cost. I show that the financial frictionscreated by this costly state verification increase the asymmetry in the movements of lendingrates, investment, and output.The basic setting of my model is