Reducing the Cost of Ex Post Bailouts with Ex Ante Regulation: Evidence From Building Codes

Transcription

1 Reducing the Cost of Ex Post Bailouts with Ex Ante Regulation: Evidence From Building Codes Tatyana Deryugina University of Illinois November 2, 2013 Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

2 Introduction The government often acts as an insurer of last resort Presence of ex post aid can theoretically create moral hazard (e.g., Kaplow 1991, Hellman et al. 2000, Raschky et al. 2007, Farhi and Tirole 2012). In theory, ex ante policies can ameliorate the moral hazard problem. One such policy is to mandate protection measures (building codes). However, the effectiveness of building codes is ultimately an empirical question. Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

3 Basic empirical approach Introduction Examine the effect of building codes on ex post disaster spending during the 2004 hurricane season in Florida. Compare aid in zip codes with stricter building codes to aid in zip codes with laxer building codes, holding constant the strength of the hurricane. Measure strictness of building code using simple wind load metric. Use plausibly exogenous variation to identify causal effects: instrument for building code strictness using historic ( ) hurricane wind speeds around the zip code. Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

4 Background and Data Building codes in the United States No national building code, states decide on their own. States increasingly adopted building codes developed by the International Code Council (ICC). Florida adopted in Hurricane-related provisions developed by the American Society of Civil Engineers (ASCE), ASCE 7. ASCE 7 also provides maps of wind speeds that a building in a given area should be able to withstand ( wind loads ). Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

5 Background and Data Figure 1: Florida Wind Loads Wind load ranges Source: Applied Technology Council, ASCE 7-98 Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

6 Data Background and Data Wind load data manually downloaded for every 0.01 degree latitude/longitude from Applied Technology Council website. Zip-level data on disaster grants to individuals ( Individual Assistance ) from FEMA for each hurricane, broken down by renters and homeowners. Actual wind speed fields from H*Wind (NOAA). Four hurricanes hit Florida in 2004: Charley (and Tropical Storm Bonnie), Frances, Ivan, and Jeanne. Sample of interest is all observations in the H*Wind dataset. Historic wind speeds (going back to 1851) from Best Tracks dataset (NOAA), matched to zip codes. Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

7 Identification strategy Results Endogeneity concern: wind loads may be correlated with unobservable determinants of damages (and thus, of federal aid). Solution: instrument for wind load using historic hurricane record. Use historic wind speeds and number of storms of various intensity of all zip codes within 500 miles. Compute weighted average using 1 distance as the weight. Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

8 Results Table: The effect of building codes on federal spending, 2SLS (1) (2) (3) (4) Panel A: Log of total aid Wind load *** * (0.006) (0.006) Log of wind load *** * (0.759) (0.799) Median year built *** *** *** *** (0.007) (0.007) (0.007) (0.007) Log of median housing value *** *** (0.150) (0.150) Observations 1,579 1,579 1,579 1,579 R-squared First-stage partial F Panel B: Log of (total aid + 1) Wind load *** *** (0.009) (0.009) Log of wind load *** *** (1.132) (1.147) Median year built *** *** (0.008) (0.008) (0.008) (0.008) Log of median housing value *** *** (0.201) (0.200) Observations 1,601 1,601 1,601 1,601 R-squared First-stage partial F Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

9 Results Table: The effect of building codes on federal spending, renters, 2SLS (1) (2) (3) (4) Panel A: Log of aid to renters Wind load ** (0.007) (0.007) Log of wind load ** (0.885) (0.916) Median year built *** *** *** *** (0.008) (0.009) (0.008) (0.009) Log of median housing value *** *** (0.158) (0.159) Observations 1,445 1,445 1,445 1,445 R-squared First-stage partial F Panel B: Log of (aid to renters + 1) Wind load (0.010) (0.010) Log of wind load * (1.220) (1.289) Median year built *** *** *** *** (0.011) (0.011) (0.011) (0.011) Log of median housing value *** *** (0.242) (0.242) Observations 1,505 1,505 1,505 1,505 R-squared First-stage partial F Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

10 Results Table: The effect of building codes on federal spending, owners, 2SLS (1) (2) (3) (4) Panel A: Log of aid to owners Wind load *** *** (0.006) (0.006) Log of wind load *** *** (0.718) (0.752) Median year built *** *** (0.006) (0.006) (0.006) (0.006) Log of median housing value *** *** (0.144) (0.144) Observations 1,566 1,566 1,566 1,566 R-squared First-stage partial F Panel B: Log of (aid to owners + 1) Wind load *** *** (0.008) (0.008) Log of wind load *** *** (0.972) (1.010) Median year built *** *** (0.007) (0.007) (0.007) (0.007) Log of median housing value *** *** (0.198) (0.197) Observations 1,588 1,588 1,588 1,588 R-squared First-stage partial F Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11

11 Conclusion Conclusion Reducing ex ante moral hazard stemming from ex post aid is a serious concern in many areas. Credible empirical evidence on the effectiveness of various policies is lacking. The role of building codes in reducing ex post aid has not been studied. Although stricter building codes significantly reduce the amount of ex post federal aid, they are not likely to be cost-effective. Tatyana Deryugina (University of Illinois) Effect of Building Codes November 2, / 11