Denis Nadolnyak (Auburn, U.S.) Valentina Hartarska (Auburn University, U.S.)

Size: px

Start display at page:

Download "Denis Nadolnyak (Auburn, U.S.) Valentina Hartarska (Auburn University, U.S.)"

Bertina Nicholson
6 years ago
Views:

1 Denis Nadolnyak (Auburn, U.S.) Valentina Hartarska (Auburn University, U.S.) 1

2 Financial markets and catastrophic risks Emerging literature studies how financial markets are affected by catastrophic risk events including those caused by climate extremes Basic conclusion: In the absence of adequate insurance markets, commercial banks and their clients are affected by catastrophic events (Garmaise and Moskowitz, 2009) Empirical evidence In developing countries climate extremes affect agricultural loan portfolios (Collier, Katchova, and Skees, 2010; Bergyz and Schraderx 2009) Tornadoes and hurricanes impact on US community banks is localized and small (Ewing, Hein and Kruse, 2005) 2

3 Motivation Agriculture is climate dependent, especially in US Southeast (less irrigation, more floods & droughts) Credit risk/loan performance in agricultural lending may be related to climate variations/extremes Farmers get production loans mainly from Farm Credit System (FCS) $161 bln in serves more (professional) farmers in the Southeast Commercial Banks $121 bln in

4 Hypothesis Agricultural loan portfolios of banks serving agricultural producers in the Southeastern U.S. are affected by inter-annual climate fluctuations (ENSO events measured by MEI) Impacts are expected through changes in repayment patterns delinquencies (loans past due days) default with realized losses (charge offs) in the future restructuring of loans? 4

5 Possible implications Results could be suggestive of Usefulness of ENSO forecasts to both farmers and banks in decision making The degree to which existing financial system and government support mechanisms (including insurance markets) and able to diversify climate related risks 5

6 ENSO has impacts on local climate in different world regions. Impacts agriculture. Source: NOAA 7

7 ENSO Impact in the Southeast Historical data and models suggest some dependency yields (agroclimate.org; Royce et al., 2011; Martinez, 2009, 2010; Baigorria et al., 2010). Aggregate impacts of ENSO on ag. production risk: ENSO events found positively associated with agricultural disaster payments in county level analysis (Nadolnyak and Hartarska, 2009). La Niña years associated with more disaster payments. Downward volatility of corn, cotton, and peanut yields in the southeast has been consistently higher in La Niña: crop insurance implications (Nadolnyak and Vedenov, 2008). 8

8 CHOOSING INDEX INTERVAL Most recent aggregate data analysis scores Multivariate ENSO Index MEI as best predictor for corn, cotton, peanut yields (Royce, Fraisse, Baigorria, 2011) Takes 1-2 weeks for the global atmosphere to respond to tropical SST anomalies (NOAA) Impact Interval must correspond to regional growing season To match available banking data, mean MEI values over March-June are used as continuous measure of ENSO signal during growing season in Southeast US 9

9 Banking Data Call Reports: quarterly commercial banks data for in AL, FL, GA, LA, MS, and SC Agricultural Bank definition: FDIC: ag. loans / total loans > 25% Federal Reserve System (Fed): ag. loans / total loans > all banks average. Better for increasing portfolio diversification Sample: ag. loans / total loans = 5.8%. Agricultural banks by FED: ag./total loans = 16.3% 473 banks in 5 states, ~1,000 annual obs. Av. annual lending of $198mil, ag. loans about $10mil 10

10 Banking Data Agricultural loans of two types: Regular crop production loans Production loans secured by farm real estate Portfolio performance by delinquency & default of Ag Production loans (also loans secured by real estate): Ag. loans overdue days (since 2001) Ag. loans overdue >90 days (since 1991) Nonaccrual ag. production loans (since 1991) Ag. production loans charge-offs (since 1984) 11

11 Summary Statistics Variable Obs Mean Std. Dev. Min Max Non performing loans backed by 2, ,408.0 real estate ('000 $) RE Loans Delinquent >30 days 2, ,627.6 RE Loans Delinquent >90 days 2, ,319.0 Real Estate charge off ('000%) 2, ,150.6 MEI Feb-August 2, Mei squared 2, JMA La Nina 2, JMA Neutral 2, JMA El Nino 2, ONI Medium El Nino 2, ONI Medium La Nina 2, ONI Neutral 2, ONI strong EL Nino 2, ONI Weak El Nino 2, ONI Weak La Nina 2, Bank Size (mln assets) 2, Land value ($ per acre) 2,996 2, , ,783.9 RE loans by FCS (billions) 2, RE Loans by banks ( billions) 2, Number of farms (millions) 2, Value of real estate (millions) 2,996 1, ,841.8 Debt/equity ratio 2, Value of crops (billion $) 2, Dummy crisis 2,

12 Reasons for default: Model Non-strategic: Idiosyncratic shock events (divorce, illness) not captured by aggregate banking data. Canceled by diversification Systemic shock events (due to weather) are reflected in aggregate commercial bank data Strategic: when value (land) < loan. To control for strategic default, use proxy for the strategic default option (change in land value). Studies show strategic defaults significant (Briggeman et al., 2009) 13

13 Estimated equation: Model D ijt = β 0 + β 1. ENSO t-1 + β 2. LandValue ijt + β 3. LandValue ijt-1 + β 3. DummyFinCrisis t-1 + γ controls ijt-1 + e ijt D ijt measures delinquency / default on ag. loans backed by farmland for bank i in state j in year t. Main variable is loans in default and nonperforming (nonaccrual) loans - that is loans borrowers cannot repay Also loans delinquent for days Bank losses not relevant for collateralized loans (farmland) 14

14 Model ENSO measures (year preceding default/delinquency): MEI averaged over late winter early summer JMA and ONI (classification borrowed from Royce, Fraisse, Baigorria, 2011) Control variables: Bank size (scale efficiency) Shares of FCS and commercial banks in lending Crop value Debt-to-Equity ratio of farm operators (demand for loans) Production subsidies, including disaster payments 15

15 Results Fixed effects regression of nonaccrual loans backed by farm land JMA La Nina *** (27.43) JMA El Nino * (53.06) ONI Medium La Nina *** (88.36) ONI Medium El Nino * (96.05) ONI Strong El Nino ** (143.9) ONI Weak El Nino (34.26) ONI Weak La Nina *** (68.47) MEI (growing season avg) 295.9*** (77.42) MEI 2 (growing season avg) *** (82.65) Disaster Payments ** (0.240) Bank Size 1.089*** 1.099*** 1.110*** 1.101*** (0.250) (0.243) (0.241) (0.253) Land Value *** *** *** *** (0.0464) (0.0536) (0.0451) (0.0422) Lagged Land Value 0.234*** 0.286*** 0.224*** 0.211*** (0.0535) (0.0620) (0.0526) (0.0505) 16

16 Results Fixed effects regression, nonaccrual loans Lag RE Loans by FCS * 59.52*** (18.48) (28.63) (18.08) (15.22) Lag RE Loans by banks *** (35.38) (34.07) (27.97) (18.75) Lag Number of Farms ** (2442) (2280) (2449) (3976) Lag Value of Real Estate *** (0.521) (0.341) (0.579) (0.406) Lag Debt/Equity Ratio *** 108.8*** 29.93* (27.65) (29.50) (33.01) (16.33) Lag Value of Crops 21.68*** *** 41.42*** (7.857) (11.87) (10.76) (10.36) Dummy Financial Crisis *** * 1181*** (271.0) (323.3) (324.3) (452.9) Mississippi (Alabama base) 231.7*** 218.0*** 222.2*** 213.7*** (27.83) (25.91) (26.37) (26.72) States Controls yes yes yes yes Constant ** (5499) (4884) (5896) (8691) Observations R-squared Number of banks Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 17

17 Results Overall, results are contrary to expectation that nonneutral years (more climate extremes) have higher defaults: Fewer defaults in both Warm and Cold Phases By JMA, $84,000/bank less in La Nina and $94,440/bank less in El Nino By ONI (weak, medium, strong): lowest defaults in Cold phase, then in Warm phase, highest in Neutral! By MEI (continuous, averaged Feb-Jun), significant inverted parabolic relationship:

18 Results Similar results found for loans overdue >30 days and > 90 days Bank losses (charge offs associated with agricultural loans backed by farmland) are lower by about $100,000 following a strong El Nino growing season for the average banks Due to ENSO persistence (memory), indexes from Fall preceding the year when loans are taken can be used to predict default rates 19

19 Results Possible explanation: government support mechanisms overcompensate + financial markets pool the risks Lower default rates associated with higher disaster payment AL disaster payments: $0.25mil in Neutral, $51mil in La Nina; $30,000/bank less in default loans Correlation b/w state-level ag. disaster payments and MEI(Feb-Jun) AL FL GA LA MS SC CORR PVAL x x x x x x Correlation b/w crop insurance subsidies and MEI(Feb-Jun) very similar, less for counter-cyclical & deficiency payments. 20

20 Results Farmers-clients of agricultural banks are affected by weather variation but government supports reverse the impacts: Overcompensation mechanism explanation consistent with Ewing, Hein and Kruse, (2005) who found that community banks serving areas affected by natural disaster are better off in the aftermath of a catastrophic event, likely due to well functioning complementary insurance markets. Controlling for strategic default is important: Land prices in the period prior to default when collateral was valued are associated with larger defaults Current land prices are associated with fewer default loans => land prices affect farmers strategic default decisions 21

21 Results Farmers leverage is associated with higher defaults Total value of crops production is associated with higher defaults.? higher level of output => lower prices => harder to repay for low income farmers? Larger farm numbers associated with more defaults in agricultural banks portfolios Higher value of farmland is associated with more loans in default perhaps due use of more marginal land in agricultural production. 22

22 Conclusion Question Posed: Are agricultural loan portfolios affected by inter-annual climate fluctuations? Result: Non-neutral years have fewer loans in defaults contrary to expectation that banks would suffer from weather extremes Explanation: Financial markets, government supports (disaster relief, countercyclical/deficiency, and crop insurance) mitigate climate related risks associated with agricultural lending in the Southeast U.S Emerging literature on non-agricultural bank lending and catastrophic risk events reports similar results & explanations 23

23 Suggestions and questions? 24

CLIMATE EFFECTS ON RAINFALL INDEX INSURANCE PURCHASE DECISIONS

CLIMATE EFFECTS ON RAINFALL INDEX INSURANCE PURCHASE DECISIONS Authors James Novak, Professor Department of Agricultural Economics and Rural Sociology Auburn University Auburn, Alabama Denis Nadolnyak,