"Option Value of Consumer Bankruptcy"

Transcription

1 "Option Value of Consumer Bankruptcy" Ethan Cohen-Cole University of Maryland - College Park Robert H Smith School of Business June 5, 2009

2 Figure 1: Bankruptcy Filings

3 Existing Explanations: Idiosyncractic shocks, such as unemployment, health shocks, and divorce. Changes in the credit market environment: Decreased transaction costs, and expansion of credit including to riskier households (supply side); Decreased costs of ling for bankruptcy, including legal and information costs, as well as social stigma (demand side).

4 Plan Introduction Asset Side Income model Real Options model Conclude

5 Existing Explanations: Idiosyncractic shocks, such as unemployment, health shocks, and divorce. Tiny e ect (Gross and Souleles, 2002; Cohen-Cole and Duygan-Bump, 2008) Large e ect (Macro Model) (Livshits, MacGee, Tertilt, 2007) Large e ect (Rhetorical) (Barron et al, 2000; Sullivan et al, 2000) Changes in the credit market environment: Decreased transaction costs, and expansion of credit including to riskier households (supply side); Decreased costs of ling for bankruptcy, including legal and information costs, as well as social stigma (demand side).

6 Existing Explanations: Idiosyncractic shocks, such as unemployment, health shocks, and divorce. Changes in the credit market environment: Decreased transaction costs, and expansion of credit including to riskier households (supply side); Tiny e ect (Empirical) (Gross and Souleles, 2002; Cohen-Cole and Duygan-Bump, 2008) Large e ect (Empirical) (Dick and Lehnert, 2009) Large e ect (Empirical) (Cohen-Cole, Duygan-Bump, and Montoriol-Garriga 2009) Large e ect (Macro Model) (Athreya, 2004, 2007) Small e ect (Macro Model) (Livshits, MacGee, Tertilt, 2007) Decreased costs of ling for bankruptcy, including legal and information costs, as well as social stigma (demand side).

7 Existing Explanations: Idiosyncractic shocks, such as unemployment, health shocks, and divorce. Changes in the credit market environment: Decreased transaction costs, and expansion of credit including to riskier households (supply side); Decreased costs of ling for bankruptcy, including legal and information costs, as well as social stigma (demand side). Large e ect (Empirical) (Gross and Souleles, 2002; Cohen-Cole and Duygan-Bump, 2008) Small e ect (Macro Model) (Athreya, 2004, 2007)

8 Summary and Today Idiosyncractic shocks, such as unemployment, health shocks, and divorce. Use exposure to risk and real options to show very large empirical e ect Explains: Why unemployment, health, divorce are quantitatively unimportant in many models Why theoretical models come to often distinct conclusions Why bankruptcy rate was so low for so long (option only valuable once) Increase in bankruptcy rate over time Consistent with large availability of credit post-bankruptcy (see Cohen-Cole et al 2009)

9 A Baseline

10 Data Data Summary

11 Credit Bureau Data US credit bureau 4 national samples: June 2003, December 2004, June 2006, December 2007; First two are large samples of about 285,780 individuals each; Latter two are HUGE samples of 27,000,000 individuals each. 1 in 9 sample of all US households with credit. Includes typical information available in a credit history on all open credit accounts. Little measurement error; Detailed geo-code info; No individual level income and employment information.

12 Census Data local unemployment rates local poulation information race, gender composition, ages local economic information povery, income, insurance coverage

13 Summary Use two models permanent income (Gottschalk and Mo tt, various) real options (value to ling) Show empirically bankruptcy time series correlated with permanent shocks bankruptcy cross-section correlated with proxy for idiosyncratic shocks

14 Personal bankruptcy Personal Bankruptcies vs Variance of Income , Variance (MG) Personal Bankruptcies Data from Mo tt/gottschalk (2002); Correlation 39%

15 Business Bankruptcy Business Bankruptcy vs Variance of Income Variance (MG) Business Bankruptcies Business lings: Correlation 72%

16 Asset Side Permanent Income Model

17 Income Model Individual income is composition of permanent and temporary components y it = µ i + ν it If uncorrelated, we can decompose variance easily Thus permanent variance is covariance of earnings between two periods (distant) with uncorrelated transitory errors. Use estimated permanent variance to calculate temporary.

18 Income Model Individual income is composition of permanent and temporary components If uncorrelated, we can decompose variance easily σ 2 y = σ 2 µ + σ 2 ν. Thus permanent variance is covariance of earnings between two periods (distant) with uncorrelated transitory errors. Use estimated permanent variance to calculate temporary.

19 Income Model Individual income is composition of permanent and temporary components If uncorrelated, we can decompose variance easily Thus permanent variance is covariance of earnings between two periods (distant) with uncorrelated transitory errors. cov (y it, y it 0) = σ 2 µ if cov (ν it, ν it 0) = 0. Use estimated permanent variance to calculate temporary.

20 Income Model Individual income is composition of permanent and temporary components If uncorrelated, we can decompose variance easily Thus permanent variance is covariance of earnings between two periods (distant) with uncorrelated transitory errors. Use estimated permanent variance to calculate temporary. σ 2 ν = σ 2 y σ 2 µ.

21 Time Series Results

22 Time Series Results

23 Can be expanded.

24 Can be expanded. Income decomposition of permanent income µ i

25 Can be expanded. Income decomposition of permanent income µ i Break into trend and innovation components: assume a random growth process such as: dµ i = α i µdt + σ i µdζ

26 Asset Side Real Options Model

27 Real Options Model Bankruptcy has a bene t and a cost Allow to progress according to geometric brownian processes (as above): db = φbdt + ωbdz (1) dc = βcdt + ηcdy (2)

28 Real Options Model Individuals choose when to le based on this joint process hitting some individual-speci c optimum 3 Path of Bankruptcy Payoff

29 Real Options Model Why is this helpful? Recall that income process is individual speci c. Model solution can provide guidance to understand implications of variation in these processes. Demographic and economic factors were not predictors of bankruptcy Can the real-options model provide inference here?

30 Real options model Figure shows 100 simulations for an economy with identical process 100 simulations with a mean-preserving spread of α = φ β

31 Real Options Model Allow individual s speci c processes to capture idiosyncratic risk db i = φ i Bdt + ω i Bdz (3) dc i = β i Cdt + η i Cdy. (4)

32 Real Options Model

33 Choosing Proxies and Groups The question is then can we capture idiosyncratic risk using higher moments of economic variables? Use unemployment across a subdivision of income and education (5 quintiles of each) local will now encompass location as well as socio-economic group

34 Choosing Proxies and Groups Bankruptcy

35 Choosing Proxies and Groups Unemployment

36 Cross-Section Results Probit: bankruptcy decision (Reports Marginal E ects at average)

37 Cross-Section Results Variance Unemployment

38 one unit increase in variance increases prob(bankruptcy) by percentage points

39 one unit increase in variance increases prob(bankruptcy) by percentage points ten percent increase in variance of the median group increases prob(bankruptcy) by ~14-20 percentage points

40 one unit increase in variance increases prob(bankruptcy) by percentage points ten percent increase in variance of the median group increases prob(bankruptcy) by ~14-20 percentage points increase median group to highest variance and increase prob(bankruptcy) by >30 percentage points

41 Cross-Section Results Kurtosis Unemployment

42 Conclusion Income risk important component in bankruptcy decision

43 Conclusion Income risk important component in bankruptcy decision "option" to le is salient and economically large

44 Conclusion Income risk important component in bankruptcy decision "option" to le is salient and economically large Policy:

45 Conclusion Income risk important component in bankruptcy decision "option" to le is salient and economically large Policy: According to model: increasing ling costs (BACPCA) only delays bankruptcy, doesn t reduce it

46 Conclusion Income risk important component in bankruptcy decision "option" to le is salient and economically large Policy: According to model: increasing ling costs (BACPCA) only delays bankruptcy, doesn t reduce it To reduce bankruptcy, reduce income risk