顯微鏡下的驟停.pdf

《顯微鏡下的驟停.pdf》由會員分享，可在線閱讀，更多相關《顯微鏡下的驟停.pdf（53頁珍藏版）》請在三個皮匠報告上搜索。

1、Sudden Stops Under the MicroscopeCecilia DassattiRodrigo LluberasCentral Bank of UruguayORT Uruguay UniversityPablo OttonelloDiego PerezMaryland&NBERNYU&NBERNovember 14,202425th Jacques Polak Annual Research ConferenceIntroduction Emerging markets experience recurrent episodes of“sudden stops”Abrupt

2、 reversals of capital flows Macro level:contractions in economic activity,consumption,aggregateinvestment,currency depreciation(e.g.,Calvo et al.06)Goal of this paper:Study the micro-level patterns of adjustmentInform theories explaining macro adjustments Approach:Exploit cross-sectional patterns in

3、 firms borrowing(Khwaja Mian 08,Di Giovanni Kalemli-Ozcan Ulu Baskaya 22)Focus:Uruguay,EM featuring 2 sudden stops and rich loan-level dataSummary1.What are the channels of transmission of sudden stops?Lender channel(e.g.,Calvo 04,Morelli Ottonello Perez 22)Collateral channel(e.g.,Korinek and Mendoz

4、a 14)Risk channel(Neumeyer Perri 05,Uribe Yue 06,Hegarty et al.24)2.Are sudden stops different from regular business cycles?Lender channel more than doubles during episodes of sudden stopsConsistent with intermediaries acceleration mechanisms(e.g.,CaballeroKrishnamurthy 01,Gertler Kiyotaki 10;He Kri

5、shnamurthy 12,Brunnermeier Sannikov 14)We do not find a strengthening of the collateral and risk channels3.Are the effects of sudden stops heterogeneous?Lender channel affects more firms with high risk and in nontradable sectorLess heterogeneity for risk and collateral channelsSummary1.What are the

6、channels of transmission of sudden stops?Lender channel(e.g.,Calvo 04,Morelli Ottonello Perez 22)Collateral channel(e.g.,Korinek and Mendoza 14)Risk channel(Neumeyer Perri 05,Uribe Yue 06,Hegarty et al.24)2.Are sudden stops different from regular business cycles?Lender channel more than doubles duri

7、ng episodes of sudden stopsConsistent with intermediaries acceleration mechanisms(e.g.,CaballeroKrishnamurthy 01,Gertler Kiyotaki 10;He Krishnamurthy 12,Brunnermeier Sannikov 14)We do not find a strengthening of the collateral and risk channels3.Are the effects of sudden stops heterogeneous?Lender c

8、hannel affects more firms with high risk and in nontradable sectorLess heterogeneity for risk and collateral channelsSummary1.What are the channels of transmission of sudden stops?Lender channel(e.g.,Calvo 04,Morelli Ottonello Perez 22)Collateral channel(e.g.,Korinek and Mendoza 14)Risk channel(Neum

9、eyer Perri 05,Uribe Yue 06,Hegarty et al.24)2.Are sudden stops different from regular business cycles?Lender channel more than doubles during episodes of sudden stopsConsistent with intermediaries acceleration mechanisms(e.g.,CaballeroKrishnamurthy 01,Gertler Kiyotaki 10;He Krishnamurthy 12,Brunnerm

10、eier Sannikov 14)We do not find a strengthening of the collateral and risk channels3.Are the effects of sudden stops heterogeneous?Lender channel affects more firms with high risk and in nontradable sectorLess heterogeneity for risk and collateral channelsOutline1.The macro dynamics during sudden st

11、ops2.Micro-level data description3.The micro dynamics during sudden stopsWhat are the channels of transmission of sudden stops?Are sudden stops different?Are the effects of sudden stops heterogeneous?4.ConclusionsThe Macro DynamicsDuring Sudden StopsSudden Stop Episodes:Uruguayan Current Account Dyn

12、amics-2.382.81-5.56-1.17-6-4-2024%1990199520002005201020152020Current account as a percentage of GDPGlobal and Regional Context during Sudden Stops2002 episode2009 episodeExcess bond premium and VIX-2024index2.533.54logs1999q42000q42001q42002q42003q4VIX(LHS)EBP(RHS)EBP EME(RHS)-2024index2.533.54logs

13、2007q12008q12009q12010q12011q1Emerging-market sovereign spreads05001000150020002500bp02000400060008000bp1999q42000q42001q42002q42003q4ARG(LHS)URU(RHS)EME excl ARG(RHS)05001000150020002500bp02000400060008000bp2007q12008q12009q12010q12011q1Banking Sector Dynamics during Sudden Stops2002 episode2009 ep

14、isode-40-2002040%-80-4004080%1999q42000q42001q42002q42003q4Non-resident Deposits(LHS)Deposits(RHS)Credit(RHS)-40-2002040%-80-4004080%2007q12008q12009q12010q12011q1Growth in Real Credit,Deposits,and Non-resident DepositsMacro Dynamics during Sudden Stops2002 episode2009 episodeReal GDP growth and une

15、mployment rate5101520%-20-1001020%1999q42000q42001q42002q42003q4RGDP growth(LHS)U rate(RHS)5101520%-20-1001020%2007q12008q12009q12010q12011q1Investment and real exchange rate depreciation-70-3503570%-60-40-200204060%1999q42000q42001q42002q42003q4Investment Growth(LHS)Real Depreciation(RHS)-70-350357

16、0%-60-40-200204060%2007q12008q12009q12010q12011q1Consumption and NERData DescriptionData Description Credit Register:comprehensive data on loans across the financial systemDetailed information on borrowers and loans:Loan amount,currency,maturity,and collateralBorrowers country of residency,sector,an

17、d credit risk category Balance sheet and income statement data for all financial institutions Coverage of combined firm-bank dataset:109,419 firms,23 banks,and 170,924 bank-firm pairs from 1999 to 2019Firms Borrowing Dynamics2002 episode2009 episodeFirms average borrowing-60-40-20020402000q12001q120

18、02q12003q12004q1Debt growth(%)RGDP growth(%)-15-10-5051015202007q12008q12009q12010q12011q1Descriptive statisticsAll firmsFirms w multiple banksFirms w multiple collateralFirms w multiple credit risklogbjtlogbjt#banks relationshipslogbjt#collateral typeslogbjtUnsecured loans(%)Mean-3.57.02.4-3.62.0-1

19、.926.8Median-5.4-1.72.0-4.52.0-3.61.6SD99.280.80.880.20.260.538.1Bottom 5%-152.9-96.52.0-124.82.0-97.80.0Top 95%161.7139.74.0122.22.099.1100.0Number of units109,4092,1232,1237,7657,76524,14424,144Observations1,919,998345,327345,32751,47451,474345,327345,327Banks Returns and Collateral during Sudden

20、Stops2002 episode2009 episodeBanks return on assets-20-1001020-20-15-10-502000q12001q12002q12003q12004q1ROA(%,LHS)RGDP growth(%,RHS)246810-.50.511.522007q12008q12009q12010q12011q1Collateral values-30-20-1001020-40-2002040602000q12001q12002q12003q12004q1Collateral growth(%,LHS)RGDP growth(%,RHS)24681

21、0-1001020302007q12008q12009q12010q12011q1Transmission ChannelsThe Lender ChannelTheories linking sudden stops to the balance sheet of financial intermediaries shocks affecting intermediaries net worth reduce the supply of credit for domestic agents(e.g.,Calvo 04,Morelli Ottonello Perez 22)Firm-level

22、 empirical model:logbjt+h logbjt1=jh+th+hZjt+hXjt1+jt+h bjt:Debt of firm j in period t Zjt:Firm-level exposure to lender channel:Average return on assets of banks linked to firm j Xjt1:Firm-level controls(NPL ratio,liquid collateral ratio,credit risk,age,.)The Lender Channel:Firm-level Estimates-.50

23、.51p.p.02468Quarterlogbjt+h logbjt1=jh+th+hZjt+hXjt1+jt+hThe Lender Channel:Firm-level Estimates-.50.51p.p.02468Quarter2002 episode2009 episodeZ(pp)Effect(pp)Share(%)Z(pp)Effect(pp)Share(%)Lender channel-17.2-10.129-1.57-1.137.8Obs.debt growth-34.7-14.5The Lender Channel:Loan-level EstimatesEmpirica

24、l model:logbijt+h logbijt1=ih+jth+hRit+hXijt1+ijt+h,bijt:Debt of firm j from bank i in period t Rit:Return on assets of banks i in period t jth:absorbs firm-specific credit demand shocks(e.g.,Khwaja and Mian,2008)The Lender Channel:Loan-level EstimatesFirm-level estimatesLoan-level estimates-.50.51p

25、.p.02468Quarter0.511.522.5p.p.02468QuarterRobustnessThe Collateral ChannelTheories linking sudden stops to changes in collateral values Negative shocks lead to contractions in collateral values,which induce deleveraging Deleveraging induces further contractions in collateral values,leading to a down

26、ward spiral(e.g.,Mendoza 02,10;Bianchi 11)Empirical models Firm level:logbjt+h logbjt1=jh+th+hZjt+hXjt1+jt+hZjt=Pkjktqkt:average change in the value of collateral associated with loans to firm j Loan level:logbkjt+h logbkjt1=kh+jth+hqkt+hXkjt1+kjt+hjthfirm-time fixed effects,khcollateral-type fixed

27、effectsThe Collateral ChannelTheories linking sudden stops to changes in collateral values Negative shocks lead to contractions in collateral values,which induce deleveraging Deleveraging induces further contractions in collateral values,leading to a downward spiral(e.g.,Mendoza 02,10;Bianchi 11)Emp

28、irical models Firm level:logbjt+h logbjt1=jh+th+hZjt+hXjt1+jt+hZjt=Pkjktqkt:average change in the value of collateral associated with loans to firm j Loan level:logbkjt+h logbkjt1=kh+jth+hqkt+hXkjt1+kjt+hjthfirm-time fixed effects,khcollateral-type fixed effectsThe Collateral ChannelTheories linking

29、 sudden stops to changes in collateral values Negative shocks lead to contractions in collateral values,which induce deleveraging Deleveraging induces further contractions in collateral values,leading to a downward spiral(e.g.,Mendoza 02,10;Bianchi 11)Empirical models Firm level:logbjt+h logbjt1=jh+

30、th+hZjt+hXjt1+jt+hZjt=Pkjktqkt:average change in the value of collateral associated with loans to firm j Loan level:logbkjt+h logbkjt1=kh+jth+hqkt+hXkjt1+kjt+hjthfirm-time fixed effects,khcollateral-type fixed effectsThe Collateral ChannelFirm-level estimatesLoan-level estimates-.1-.050.05.1.15p.p.0

31、2468Quarter0.1.2.3.4.5.6p.p.02468QuarterRobustnessTransmission Channels:Quantifying Aggregate Effects2002 episode2009 episodeZ(pp)Effect(pp)Share(%)Z(pp)Effect(pp)Share(%)Lender channel-17.2-10.129-1.57-1.137.8Collateral channel-58-5.716.3-12.6-1.248.5Obs.debt growth-34.7-14.5The Risk ChannelTheorie

32、s linking sudden stops to changes in external risky borrowing costs(e.g.,Neumeyer Perri 05,Uribe Yue 06,Hegarty et al.,22)Increases in the global price of risk raise borrowing costs for risky agents,reducing their borrowingEmpirical models Firm level:logbjt+h logbjt1=jh+th+hZjt+hXjt1+jt+hZjt=Riskjt1

33、EBPt,Riskjt1:firm js share of unsecured debt Loan level:logbijt+h logbijt1=ih+jth+hZijt+hXijt1+ijt+hZijt=Riskijt1EBPt,Riskijt1:dummy=1 if debt unsecuredjthfirm-time fixed effects,ihsecured/unsecured fixed effectsThe Risk ChannelTheories linking sudden stops to changes in external risky borrowing cos

34、ts(e.g.,Neumeyer Perri 05,Uribe Yue 06,Hegarty et al.,22)Increases in the global price of risk raise borrowing costs for risky agents,reducing their borrowingEmpirical models Firm level:logbjt+h logbjt1=jh+th+hZjt+hXjt1+jt+hZjt=Riskjt1EBPt,Riskjt1:firm js share of unsecured debt Loan level:logbijt+h

35、 logbijt1=ih+jth+hZijt+hXijt1+ijt+hZijt=Riskijt1EBPt,Riskijt1:dummy=1 if debt unsecuredjthfirm-time fixed effects,ihsecured/unsecured fixed effectsThe Risk ChannelTheories linking sudden stops to changes in external risky borrowing costs(e.g.,Neumeyer Perri 05,Uribe Yue 06,Hegarty et al.,22)Increase

36、s in the global price of risk raise borrowing costs for risky agents,reducing their borrowingEmpirical models Firm level:logbjt+h logbjt1=jh+th+hZjt+hXjt1+jt+hZjt=Riskjt1EBPt,Riskjt1:firm js share of unsecured debt Loan level:logbijt+h logbijt1=ih+jth+hZijt+hXijt1+ijt+hZijt=Riskijt1EBPt,Riskijt1:dum

37、my=1 if debt unsecuredjthfirm-time fixed effects,ihsecured/unsecured fixed effectsThe Risk Channel(e)Firm-level estimates(f)Loan-level estimates-.04-.03-.02-.010p.p.02468Quarter-3-2-101p.p.02468QuarterSimilar results using the VIX,as in Di Giovanni,Kalemli-Ozcan,Ulu,and Baskaya(2022)DetailsTransmiss

38、ion Channels:Quantifying Aggregate Effects2002 episode2009 episodeZEffect(pp)Share(%)ZEffect(pp)Share(%)Lender channel-17.2 p.p.-10.129-1.6 p.p.-1.17.8Collateral channel-58 p.p.-5.716.3-12.6 p.p.-1.28.5Risk channel0.9 s.d.-.641.94.9 s.d.-5.840Obs.debt growth-34.7-14.5Are sudden stops different?Are S

39、udden Stops Different?Empirical models:logbijt+h logbijt1=ih+jth+hZijt+hZijtSt+hXijt1+ijt+hLender channelCollateral channelRisk channel0.511.522.5p.p.02468Quarter-1-.50.5p.p.02468Quarter-10-505p.p.02468QuarterFirm-level estimatesRobustnessAre Sudden Stops Different?Empirical models:logbijt+h logbijt

40、1=ih+jth+hZijt+hZijtSt+hXijt1+ijt+hLender channelCollateral channelRisk channel0.511.522.5p.p.02468Quarter-1-.50.5p.p.02468Quarter-10-505p.p.02468QuarterFirm-level estimatesRobustnessAre the effects of sudden stopsheterogeneous?Heterogeneous Effects:The Lender ChannelImpactPeakAverageObsABy type of

41、collateralUncollateralized.233.81.676,211(.423)(1.61)Collateralized.251.9.93311,100(.135)(.773)BBy firms riskLow risk-.171.1.4218,505(.16)(.957)High risk.92.61.5169,909(.259)(1.08)CBy firms sectorTradable.11.88.23123,235(.227)(.964)Non-tradable.32.41.2435,626(.137)(.668)DBy currency denominationLoca

42、l currency-.823.31.5209,533(.33)(1.23)Foreign currency.421.3.61382,719(.115)(.638)EBy type of bankLocal banks.4762.763,041(.143)(1.42)Foreign banks.122.21.5245,086(.439)(.944)Private banks-.0631.4.64338,420(.133)(.693)Excluding failed banks.561.81.2539,072(.235)(.559)Heterogeneous Effects:The Collat

43、eral ChannelImpactPeakAverageObsABy firms riskLow risk.62.62.375,558(.043)(.043)High risk.6.6.4112,468(.031)(.031)BBy firms sectorTradable.66.66.343,552(.046)(.046)Non-tradable.58.58.4314,474(.031)(.031)CBy currency denominationLocal currency.61.61.44,638(.047)(.047)Foreign currency.61.61.3813,422(.

44、028)(.028)DBy type of bankLocal banks.83.96.412,573(.079)(.31)Foreign banks.68.68.476,076(.045)(.045)Private banks.63.63.428,637(.033)(.033)Excluding failed banks.61.61.3916,860(.027)(.027)Heterogeneous Effects:The Risk ChannelImpactPeakAverageObsABy firms sectorTradable-.011-.06-.026123,235(.01)(.0

45、2)Non-tradable-.0086-.0119-.0091435,626(.0051)(.017)BBy currency denominationLocal currency-.015-.035-.013209,533(.012)(.021)Foreign currency-.0009-.0035.0049382,719(.005)(.0088)CBy type of bankLocal banks-.023-.025-.01463,041(.012)(.029)Foreign banks-.0079-.0256-.0142245,086(.0083)(.0163)Private-.0

46、09-.0293-.0132338,420(.0063)(.0127)Excluding failed banks-.01-.025-.016539,072(.0046)(.0093)ConclusionsConclusions Evidence transmission channels of sudden stops at the micro level Findings highlight two central factors in these episodes:Intermediaries balance sheets:distinguishing feature setting s

47、udden stopsapart from regular business cyclesFirms default risk,which amplifies these effects Policies targeting financial stability,firms indebtedness,and bankruptcyresolution can be central to mitigating the effects of sudden stops Future research:Study these policies by combining empirical estima

48、tes withquantitative models of sudden stopsThank you!ChronologyTable 1:2002 Sudden Stop Chronology8/1998Russias devaluation1/1999Brazils devaluation6/2001Uruguay extends crawling exchange rate band12/2001Argentina establishes the“Corralito”Argentinean president quitsIMF suspends loan disbursements t

49、o ArgentinaArgentina defaults1/2002Argentinas devaluationUruguay re-extends crawling exchange rate band2/2002Argentina establishes the“Corral on”BCU intervenes to re-capitalize Banco Comercial,the main private bankBCU suspends Banco Galicia Uruguay and bank run startsUruguay loses investment grade s

50、tatusUruguays Congress approves fiscal adjustment package3/2002Agreement between IMF and Uruguay was achieved for$743 million4/2002Foot-and-mouth disease(“Aftosa”)breaks in Uruguay5/2002Uruguays Congress approves 2nd fiscal adjustment package6/2002BCU intervenes in Banco Montevideo and La Caja Obrer

51、aUruguay abandons crawling peg exchange rate regime7/2002Uruguays Minister of Economy quits followed by the president of the BCUUruguay declares bank holiday8/2002Bank holiday lifted after$3 billion bailout package provided by the US and mulilateralsFiscal and Reserves dynamicsFigure 1:Fiscal and Re

52、serve Dynamics during Sudden Stop Episodes2002 episode2009 episodea)Consolidated Fiscal Balance-3000-2000-1000010002000million pesos1999q42000q42001q42002q42003q4-3000-2000-1000010002000million pesos2007q12008q12009q12010q12011q1b)Real Government Spending Growth-15-10-5051015%1999q42000q42001q42002q

53、42003q4-15-10-5051015%2007q12008q12009q12010q12011q1c)Official Reserve Assets50020003500500065008000million USD1999q42000q42001q42002q42003q450020003500500065008000million USD2007q12008q12009q12010q12011q1Figure 2:Economic Activity during Sudden Stop Episodes2002 episode2009 episodeb)Investment and

54、Consumption Growth-30-1501530%-60-40-200204060%1999q42000q42001q42002q42003q4Investment(LHS)Consumption(RHS)-30-1501530%-60-40-200204060%2007q12008q12009q12010q12011q1c)Nominal and Real Depreciation-20020406080100%-20020406080100%1999q42000q42001q42002q42003q4NominalReal-20020406080100%-200204060801

55、00%2007q12008q12009q12010q12011q1BackFirms borrowing during Sudden StopsFigure 3:Firms Borrowing during Sudden Stops2002 episode2009 episode(a)Firms borrowing-60-40-200202000q12001q12002q12003q12004q1All firmsMultibank Firms-10-50510152007q12008q12009q12010q12011q1All firmsMultibank Firms(b)Local cu

56、rrency debt-60-40-200202000q12001q12002q12003q12004q1All firmsMultibank Firms-20-100102007q12008q12009q12010q12011q1All firmsMultibank Firms(c)Foreign currency debt-60-40-200202000q12001q12002q12003q12004q1All firmsMultibank Firms-10-505102007q12008q12009q12010q12011q1All firmsMultibank FirmsFigure

57、4:Average dynamics of firms collateral(liquid assets)Gov.debt instrumentCorp.debt instrument2002200920022009-30-20-100102030-100-500501002000q12001q12002q12003q12004q1246810-20020402007q12008q12009q12010q12011q1-20-1001020-500501001502000q12001q12002q12003q12004q1246810-2002040602007q12008q12009q120

58、10q12011q1CashPublic credit guarantee-20-1001020-1001020302000q12001q12002q12003q12004q1246810-20-10010202007q12008q12009q12010q12011q1-20-1001020-20-10010202000q12001q12002q12003q12004q1246810-10010202007q12008q12009q12010q12011q1Figure 5:Average dynamics of firms collateral(illiquid assets)Real es

59、tateVehicles2002200920022009-20-1001020-100102030402000q12001q12002q12003q12004q1246810-1001020302007q12008q12009q12010q12011q1-20-1001020-1001020302000q12001q12002q12003q12004q1246810-100102030402007q12008q12009q12010q12011q1Goods/MerchandiseCattle-20-1001020-100-500501002000q12001q12002q12003q1200

60、4q1246810-40-20020402007q12008q12009q12010q12011q1-20-100102001020302000q12001q12002q12003q12004q1246810-10010202007q12008q12009q12010q12011q1Figure 6:Change in asset prices and change in collateral value2002 episode2009 episode(a)Cattle-.20.2.42000q32001q32002q32003q3CollateralPriceCattle-.6-.4-.20

61、.2.42007q12008q12009q12010q12011q1CollateralPriceCattle(b)Real Estate-.3-.2-.10.1.22000q12001q12002q12003q12004q1CollateralPriceReal estate-.2-.10.1.22007q12008q12009q12010q12011q1CollateralPriceReal estate(c)Land-1-.50.52001q32002q12002q32003q12003q3CollateralPriceLand-.2-.10.1.2.32007q12008q12009q

62、12010q12011q1CollateralPriceLandThe risk channel-Robustness using VIX(a)Firm-level estimates(b)Loan-level estimates-.06-.04-.020p.p.02468Quarter-4-3-2-10p.p.02468Quarter-.02-.010.01.02p.p.02468Quarter-4-20246p.p.02468QuarterAgg effects of the risk channel-Robustness2002 episode2009 episodeZ(pp)Effec

63、t(pp)Share(%)Z(pp)Effect(pp)Share(%)Lender channel-17.2-3.49.7-1.57-0.382.6Collateral channel-58-1.75-12.6-0.382.6Risk channela0.86-0.381.124.9-3.4824Observed credit growth(bp)-3,470-1,449Notes:This table reports back-of-envelope calculations for the different channels,usingfirm-level estimates.The

64、first column shows the change in Z,while the second column showsthe effect computed asZ,where correspond to the average effect within 8 quarters.Thethird column correspond to column(2)as a%of the observed credit growth.a:Z for therisk channel correspond to changes in EBP(std)times the share of unsec

65、ured debt.Are Sudden Stops different?Figure 7:Channels of Transmission:Differential Effects During Sudden Stops-Firm levelestimatesLender channelCollateral channelRisk channel-10123p.p.02468Quarter-.20.2.4.6p.p.02468Quarter-.06-.04-.020.02p.p.02468QuarterBackTable 2:Channels of Transmission:Robustne

66、ss AnalysisImpactPeakAverageObsPanel A:Lender ChannelABaseline.231.8.91558,861(.117)(.559)BRobustnessNo loan-level controls.251.8.96558,861(.117)(.561)Separate firm and time FE.321.5.77558,861(.098)(.441)Panel B:Collateral ChannelABaseline.61.61.418,164(.025)(.025)BRobustnessNo loan-level controls.6

67、1.61.418,164(.025)(.025)Separate firm and time FE.64.64.3418,164(.02)(.02)Panel C:Risk ChannelABaseline-.9-2.1-1.3558,861(.454)(.915)BRobustnessNo loan-level controls-1-2.3-1.6558,861(.454)(.917)Separate firm and time FE-.4-1.3-.32558,861(.348)(1.09)Lender channelCollateral channelRisk channelTable

68、3:Differential Effects of Channels of Transmission:Robustness AnalysisImpactPeakAverageObsPanel A:Lender ChannelABaseline.681.91.4558,861(.144)(.392)BRobustnessShorter sudden stop window.293.11.9558,861(.148)(.508)CBy Sudden stop episode2002 episode.38.63.27558,861(.17)(1.25)2009 episode.992.21.7558

69、,861(.175)(.411)Panel B:Collateral ChannelABaseline.018.14-.04818,164(.056)(.183)BRobustnessShorter sudden stop window.048.16.0009718,164(.065)(.249)CBy sudden stop episode2002 episode.009.29.009218,164(.065)(.262)2009 episode.037.097-.10518,164(.092)(.342)Panel C:Risk ChannelABaseline.6-4.3-.84558,861(.999)(2.05)BRobustnessShorter sudden stop window.29-4.7-.76558,861(1)(2.05)CBy sudden stop episode2002 episode4.3-2.38.8558,861(2.2)(18)2009 episode-.047-4.8-1.4558,861(1.1)(2.1)Back