Abuse of Dutch insolvency law: strategic defaulting on mortgages increased due to the 2008 financial crisis

Abuse of Dutch Insolvency Law: Strategic

Defaulting on Mortgages Increased Due to the

2008 Financial Crisis

Amsterdam, July 1st_{, 2014}

Bachelor Thesis by Fouad Ben Masoud Supervised by J.C.M van Ophem

Abstract

Strategic default behavior suggests that the default process is not only a matter of inability to pay. Economic and social cost also contribute to the likelihood of strategic default. Unlike prior research, no conclusive evidence was found that strategic defaulting has increased post the financial crisis. Despite that significance was found in the increase of debt, this, however, could not be related to strategic defaulting on mortgages with negative equity. The main finding of this paper is that the cooperativeness of the debtor is the major determinant that determines the outcome of the amicable process, thus creating a wrong incentive for the debtor.

Table of Contents

1. INTRODUCTION ... 4

2. REVIEW OF LITERATURE ... 6

2.1STRATEGIC DEFAULTING AND DECISION-MAKING ... 6

2.2MORAL HAZARD ... 11

3. METHODOLOGY & DATA ... 14

3.1METHODOLOGY DESCRIPTION ... 14

3.2DATA DESCRIPTION ... 15

3.3HYPOTHESES ... 16

3.4HYPOTHESES TESTING ... 16

3.5INTERPRETATION OF THE COEFFICIENTS ... 17

4. RESULTS & ANALYSIS ... 18

4.1DESCRIPTIVE STATISTICS ... 18

4.2UNRESTRICTED PROBIT REGRESSION ... 19

4.3RESTRICTED MODEL ... 23

4.4LIKELIHOOD RATIO TEST ... 26

5. DISCUSSION AND CONCLUSION ... 27

6. BIBLIOGRAPHY ... 28

1. Introduction

The financial crisis of 2008 and the subsequent European sovereign debt crisis caused a severe pressure and impairment on the Dutch economy. Affecting the unemployment rate to a new peak of 8.7 percent, i.e. 687 unemployed.1 _{The same} adverse development is witnessed in the real housing prices, where a decline of 20 percent is observed compared to 2008.2 _{The immobility of homeowners to sell} their house is aggravated by the loan to value ratio (hereafter LTV), 34 percent of the mortgagees encounter negative equity in 2014, which amounts to 1.4 million households,3 _{hence selling the property is usually not considered as a desired} option. Of the 400 thousand houses that are for sale, 47 percent are lingering for a period of over 1.5 years.4 _{In addition, in 2013, there were 6.1 percent fewer} houses sold than in the previous year.5

Gradually more households encounter difficulties meeting their financial liabilities. The Dutch equivalent of the Bureau Debt Registration (BKR) has registered 91.812 households in 2013 whose mortgages are due for over 120 days.6 _{The debt aid program, called the amicable process (Dutch: minnelijk} traject), seems to be unable to address the overdue payments in the case of mortgages. The amicable process counselors negotiate between debtors and creditors. Once full redemption is regarded as impossible, they are allowed to offer four debt settlements: a payment plan, debt negotiation, consolidation of loans, or refinancing the debt. These settlements are limitative by law and only in effect once all creditors concede. If these attempts are unavailing for all parties, the counselor forwards the application towards the court, which then decides if

1 _{“CBS labor market review April 2014,” last modified June 3, 2014,}

http://www.cbs.nl/NR/rdonlyres/65EB9989-B24E-4108-8F9D-2A20999FBC3C/0/pb14n033.pdf

2 _{“CBS Dutch Housing Prices, Mortgage and Consumption Rate,” last modified June 3, 2014,}

http://www.cpb.nl/publicatie/de-nederlandse-woningmarkt-hypotheekrente-huizenprijzen-en-consumptie

3 _{“CBS Disposable Income Review,” last modified June 3, 2014}

http://www.cbs.nl/nl-NL/menu/themas/inkomen-bestedingen/publicaties/artikelen/archief/2014/2014-4044-wm.htm

4 _{“ CBS Constructing and Housing,” last modified June 3, 2014, http://www.cbs.nl/nl-}

NL/menu/themas/bouwen-wonen/publicaties/artikelen/archief/2014/2014-4075-wm.htm

5 _{“Statline CBS selling prices 1995-2013,”last modified June 3, 2014}

http://statline.cbs.nl/StatWeb/publication/?DM=SLNL&PA=81885NED&D1=3,5&D2=0,5-20&D3=4,29,54,75-94&HDR=G1,T&STB=G2&VW=T

6 _{“Debt Registration Rate Mortgages,” last modified June 3, 2014,}

http://perskamer.bkr.nl/bkr-hypotheekbarometer-bijna-10000-consumenten-met-een-achterstand-erbij/

the debtor is allowed to participate in the WSNP (Law Debt restructuring Natural Person).7 _{This strict regime subjects the debtor to provide maximum effort during} a timespan of thirty-six months, up to sixty months. The judge decides afterwards whether the participant has shown sufficient effort, to the best interest of the creditors, during this period, to grant the debtor a fresh start with a clean slate, thus exempting all remaining debt.8

With the increase of negative home equity on mortgages, mortgagors face a moral hazard risk in where potential default costs are carried by them. The WSNP program allows the mortgagees to remit all debt, after a relative short time period, including the negative home equity. Hence, this uncommon, sudden decline in housing prices would encourage mortgages with negative home equity to strategically default on their debt. This paper will investigate the notion of strategic defaulting in the Netherlands by comparing two independent samples, respectively from 2008 and 2014, of debtors who participated in the amicable process. By examining their options and setting restrictions we create a simple model that might explain opportunistic behavior. The hypothesis is that the most recent sample will exhibit a higher fail for the participants with a mortgage, since they have the most to gain. The research will be conducted by examining in section two the existing literature concerning strategic defaulting in the United States, and operationalizes the theories to make it compatible with the Dutch standards. The third section provides a more elaborated insight of the empirical methodology and date choice, followed by the results and analysis of the tests conducted in section four. Finally, in the last section a summary of the results and discussion regarding policy recommendation will be provided for future research.

7 _{Article 343 section 2 FW (Bankruptcy Law)} 8 _{Article 352 section 1 FW (Bankruptcy Law)}

2. Review of Literature

The present paper relates to the Dutch natural person bankruptcy law. Throughout the literature review the strong assumption is made that insolvent natural persons will declare for bankruptcy, while in fact most insolvent natural persons in the Netherlands default without filling for bankruptcy, as figure one exhibits. This assumption is a critical factor that enables comparison between the United States consumer bankruptcy theories with the Dutch one. The difference can be explained due to a different law, i.e. non-bankruptcy law in the United States does not provide sufficient options to force consumers to honor the initial obligation.9 _{While in the Netherlands a curator is assigned to a debtor, in the} WSNP program, to liquidize all assets without filling for bankruptcy.10

The first section of the literature study is derived from the United States bankruptcy laws Chapter 7 and 13. After which a clear demarcation will be made in the moral hazard subsection when referring the components back to the Dutch standards.

Figure 1: Development of Insolvency in the Netherlands

Source: CBS Statline

2.1 Strategic Defaulting and Decision-Making

Advocates of traditional natural person bankruptcy model assign the rise in bankruptcy fillings to economic distress. By arguing that a direct relationship

9 _{Richard M. Hynes, “Why (Consumer) Bankruptcy,” Alabama Law Review 56:1:12 (2004): 122} 10 _{Article 316 section 1 FW (Bankruptcy Law)}

0 2000 4000 6000 8000 10000 12000

Y'08 Y'09 Y'10 Y'11 Y'12 Y'13

WSNP Applicants

Clean Slate Agreement Bankruptcy Other

exists between consumer indebtedness and rising (involuntarily) bankruptcy, caused by exogenous factors.11 _{These factors are interpreted as advert events} such as (temporarily) unemployment, divorce and health problems.12 Han and Li (2011) compared the bankruptcy fillers with the non-fillers and noticed a significant difference in vulnerability to adverse events between these two groups, making them more likely to file for bankruptcy. The former tends to borrow more, spend beyond their means, have higher leverage on their collateralized loans, and are willing on paying higher premiums for their credits.1314 _{There are numerous factors that correlate with household insolvency,} and which influence households’ decision-making process relevant to the likelihood of strategic defaulting. The likelihood increases with education and non-whites.15 _{For women, the presence of children, and being a single parent.}16 Household size,17 _{health problems,}18 _{business failure,}19 _{income reduction,}20 _and increased debt,21 _{are all considered as the main drivers behind insolvency. The} likelihood however, decreases with employment,22 _{homeownership,}23 _{and age.}24

11 _{Teresa A. Sullivan, Elizabeth Warren and Jay Westbrook, The Fragile Middle Class: Americans}

Indebt (Connecticut: Yale University Press, 2000).

12 _{David Himmelstein et al., “Illness and Injury as Contributors to Bankruptcy,” Health Affairs (2005):}

63–73.

13 _{Song Han and Geng Li, “Household Borrowing after Personal Bankruptcy,” Journal of Money,}

Credit and Banking 42:2-3 (2011): 514.

14 _{Ning Zhu, “Household Consumption and Personal Bankruptcy,” The Journal of Legal Studies 40:1}

(2011): 3.

15 _{Ziegel, J. S. (2001). A Canadian Perspective The Fragile Middle Class: Americans in Debt. By}

Teresa A. Sullivan, Elizabeth Warren. Tex. L. Rev., 79, 1244

16 _{Warren, E. (2003). Growing Threat to Middle Class Families, The. Brook. L. Rev., 69, 401.}

17 _{Calem, P. S., & Mester, L. J. (1995). Consumer behavior and the stickiness of credit-card interest}

rates. The American Economic Review, 1349

18 _{Domowitz, I., & Sartain, R. L. (1999). Determinants of the consumer bankruptcy decision. The}

Journal of Finance, 54(1), 415

19 _{Warren, E. (2003). Growing Threat to Middle Class Families, The. Brook. L. Rev., 69, 401.}

20 _{Fay, S., Hurst, E., & White, M. J. (2002). The household bankruptcy decision. American Economic}

Review, 732

21 _{Godwin, D. D. (1999). Predictors of households’ debt repayment difficulties. Financial Counseling}

and Planning, 10(1), 69

22 _{Domowitz, I., & Sartain, R. L. (1999). Determinants of the consumer bankruptcy decision. The}

Journal of Finance, 54(1), 404

23 _{Ibid., 389.} 24 _{Ibid., 203.}

Zywicki (2005), however, has argued that the predictive power of the traditional model is incapable of explaining the rise in bankruptcy filling rates of the past twenty-five years.25 _{He found little empirical evidence that both equity} and bankruptcy insolvency has increased in the past twenty-five years, while the net wealth of households has reached new records.26 _{Furthermore, he argues that} neither the frequency nor severity of the financial chocks has changed substantial. Zywicki (2005) concludes that the rise in natural person bankruptcy rate is not a consequence of economic distress, but rather a way for natural persons to refuse subduing the spending rate or making an appeal on their accumulated wealth.27 Thus instead of making effort to repay creditors, the first solution would be to file for bankruptcy. While the core objective of bankruptcy filling is to create an incentive for natural persons to ensure their work performance in a situation where the benefits of working hard are reduced by the debt burden.28 _The possibility for bankruptcy filling for natural persons is considered as a necessary social insurance to withstand insoluble financial problems. Thus creating a trade-off by oppressing another key factor, that advocates personal bankruptcy protection, by providing consumption protection.29

The primary driver to strategically file for bankruptcy is defined as the debt exemption minus the level of garnishment that must be liquidated.30 _Other economic reasons are the relocation cost, the willingness to be marked with a negative credit rating and the chance of finding an alternative rental house with lower mortgage payment. Guiso, Sapienza, and Zingales (2009), estimated that if negative equity is lower than 10% of the absolute value of their houses, homeowners are unwilling to strategically default on their mortgages.31 _{Once that} threshold is surpassed, they concluded that strategic defaulting increases at a

25 _{Todd J. Zywicki, “An Economic Analysis of the Consumer Bankruptcy Crisis,” Northwestern}

University Law Review 99:4 (2005): 1463.

26 _{Ibid., 1539.} 27 _{Ibid., 1539.}

28 _{Igor Livishits, James Macgee and Michele Tertilt, “Consumer Bankruptcy: A Fresh Start,” The}

American Economic Review 97:1 (2007):402-418.

29 _{Michelle J. White, “Abuse or protection: Economics of bankruptcy reform under BAPCPA,”}

American Law & Economics Association (2006): 8.

30 _{Scott Fay, Erik Hurst and Michelle J. White, “The Household Bankruptcy Decision,” American}

Economic Review, 92(3) (2002):706-718.

31 _{Guiso, L., Sapienza, P., & Zingales, L., “Moral and social constraints to strategic default on}

mortgages,”National Bureau of Economic Research.(2009): 21.

ascending rate up to 17 percent.32 _{Fay, Hurst, and White (2002) concluded in their} research that, all else being equal, natural persons are more likely to file for bankruptcy if they live in a state with higher bankruptcy rates, suggesting the presence of spillovers.33 _{Which corresponds with the work of Gross and Souleles,} they refer to this phenomenon as the stigma effect.34 _{This implies that the social} norm regarding bankruptcy becomes more acceptable in the local society, lowering the threshold for filling bankruptcy and becoming a contagious effect. Another determining factor that lowers the threshold for filling bankruptcy are the direct legal cost and information transparency of the process. 35 _{When too low,} households take advantage of the easy access to discharge all of their debts, hence relocating the cost of default to the entire society.

A perfect rational assessment is considered as seldom probable due to the lack of full availability of information, cognitive limitation of the actor, and time. Thus current theory refers to bounded rationality.36 _{A prominent characteristic of} rational decision-making is that actors exhibit consistent behavior. 37 _The decision-making process generally consists of behavior that displays cost-benefit analysis; one chooses the alternative that maximizes the target(s) with the lowest possible cost. Aside of the financial aspects, the immaterial perspective is also taken into consideration, i.e. trust or quality of the service. Due to continuously changing circumstances and availability of new information, reassessment of alternatives may always be there if interim changes occur in the cost or benefit. An important facet is that actors might exhibit strategic behavior.38 _Assembling the bounded rationality with strategic defaulting and mortgage payment together, we can create a two period model, assigning the good period probability α and the bad period probability (1- α). In the good period the house will be worth

P

₂G whereas in the bad state, the house is worth

P

₂B. Similar to the work of Foote,

32 _{Ibid., 21.}

33 _{Fay, Hurst and White, “The Household Bankruptcy Decision,” 716.}

34 _{Gross, D. B., & Souleles, N. S., “An empirical analysis of personal bankruptcy and delinquency,”}

Review of Financial Studies 15:1 (2002): 319-347.

35 _{Han and Li, “Household Borrowing after Personal Bankruptcy,” 35.} 36 _{Simon, H. A. (1957). Models of man; social and rational.}

37 _{Dietz, F., “Rationaliteit in economie en beleid,” (Amsterdam: Boom, 1998), 160.} 38 _{Ibid., 164.}

Kristopher, and Paul,39 _{we assume}

_P

2B

< M

2

– C <

P

₂G, where M2 is the nominal

mortgage in period two and C the cost of default in period two. Which expresses the value of staying in home as

Rent1− Mpay1+ _{1 + r �α� P}1 2G − 𝑀𝑀2� − (1 − α)�P2B+ C�� + 𝜉𝜉 (2.1)

By making mortgage payment Mpay1, a mortgagee can benefit by strategically defaulting if the present value of the expected return in the second period (with mortgage rate denoted as r) plus the stigma effect (𝜉𝜉), is smaller than what he would pay for rent now. Putting the cost and benefit together, mortgagees default if and only if:

−𝐶𝐶 > Rent1− Mpay1+ _{1 + r [α� P}1 2G − 𝑀𝑀2� − (1 − α)�P2B+ C� + 𝜉𝜉 (2.2)

Rearranging the terms we obtain

Mpay1− Rent1> (r + α)C + α�P2 G _{– 𝑀𝑀}

2�

1 + r + 𝜉𝜉 (2.3) Which indicate that the cost of default (C), the discount rate (r), probability assigned to an increase of housing prices (α), the capital gain in period 2 (P2G -

M2), and the stigma effects (𝜉𝜉) are the thresholds that determine whether the premium to stay (Mpay1− Rent1) is sufficient. It should be noted that this simple

model does not take some crucial aspects into consideration, i.e. transaction cost and the mortgage interest deduction are left out of the equation.

Equation 2.3 is designed as a complementary extension of the estimated probit model in section IV, to determine whether the significant variables are affiliated with the theory. According to the theory we would expect that the determining factor on defaulting is the negative equity that households face on their mortgages. Once the present value of the negative equity increases the threshold of the stigma effect and the cost of default, households are more likely to default on their debt. Nevertheless, equation 2.3 is considered as limited and not applicable with the Dutch law, since asset and income exemptions and garnishments are left out. Ergo, the upcoming section provides further elaboration on the theory by including moral hazard in the equation.

39 _{Foote, C. L., Gerardi, K., & Willen, P. S., “Negative equity and foreclosure: Theory and evidence,”}

Journal of Urban Economics 64:2 (2008): 240.

2.2 Moral Hazard

The Dutch bankruptcy and insolvency law is, unlike the United States, a debtors’ favored law, which grants priority to their interest and position.40 _{The amicable} process objective is that all creditors should voluntarily align their different interests in finding a suitable agreement for all parties. If this is not reached, the creditors risk losing a substantial amount on their initial principal in the WSNP program. Table 1 exhibits both an increase of insolvency and success rate in finding an agreement. In juxtaposition to the doubled insolvency request in 2013 we moreover notice an ascending success rate since 2006. This can be explained due to a different more efficient approach in guiding new amicable requests set by the NVVK (Dutch Association Peoples Credit).41 _{In conjunction with the new} guidelines, the Dutch law forces the creditor to respond cooperatively in the amicable process, and leaving the debtor without an incentive. The debtor is aware of the fact that the court may not include the preferences of the creditor, in the terms of agreement of the final remediation plan, in its verdict. This remediation plan states the obligation of the debtor, a timeline, and the amount of the future income that is exempt from garnishment. Aside of the essential loss in negotiation power of the creditors, Jungmann (2006) concludes that creditor retrieve significantly less of their principal once the debtor is enrolled in the WSNP program.42

40 _{Jungmann, N., “De Wsnp: bedoelde en onbedoelde effecten op het minnelijk traject,” (Amsterdam:}

University Press, 2006), 51.

41 _{“Annual Report NVVK 2013,” last modified June 3, 2014,}

http://www.nvvk.eu/images/pdfs/Jaarverslag%20NVVK%202013.pdf

42 _{Jungmann, De Wsnp: bedoelde en onbedoelde effecten op het minnelijk traject wsnp, 66.}

Table 1: number of amicable arrangements in the period 1999-2013 Period Process Request Amicable Agreements Amicable Success rate

Y'99 28,969 7,983 28% Y'00 27,599 7,022 25% Y'01 23,890 5,973 25% Y'02 28,551 5,763 20% Y'03 34,500 5,300 15% Y'04 39,000 3,500 9% Y'06 46,000 7,820 17% Y'07 47,500 9,975 21% Y'08 44,100 14,553 33% Y'09 53,000 16,430 31% Y'10 77,000 51,590 67% Y'11 76,000 60,800 80% Y'12 84,000 52,920 63% Y'13 89,000 54,290 61%

Source: NVVK Annual reports

Assuming that a debtor with negative home equity only has three alternatives: (1) don’t default and repay all of the creditors in full, (2) default on house and repay a significant amount in the amicable process, or (3) default on house and file for the WSNP program. The determinants that influence the debtors’ decision depends on the level of garnishment permitted, the exemptions, the value of his home and the value of his human capital. In comparison to the work of Berkowitz and Hynes (1999), these determinants result in the following function: F(g, Xp, h, k).43 _{Where g is the fraction of future income} that is exempted from garnishment, Xp is the personal property exemptions, h is the value of the home, and k is the debtors’ human capital (future income). If the value of the mortgage exceeds the value of the home, an insolvent debtor would only strategically default on his mortgage if:

M -

𝜉𝜉 < h - v < M

(2.4) With the restriction:

gk + X

p

> M

(2.5)

Where M is the amount of mortgage loan, 𝜉𝜉 is the stigma effect, and v the subjective house value.

43 _{Berkowitz, J., & Hynes, R. (1999). Bankruptcy Exemptions and the Market for Mortgage Loans*.}

The Journal of Law and Economics, 42(2), 809-830.

If these conditions hold, the debtor would be inclined to default on his mortgage either by the amicable process (alternative two) or participating three years in the WSNP program (alternative three). The determinant factor depends on the amount of debt (D). The debtor would only participate in the WSNP program if

D > 3*(gk) + Xp, (2.6) The debtor improves his outlook, wherein g and k are annualized, if the principal is larger than he total amount that can be accumulate in three years to repay the creditors. Allowing him a clean slate afterwards with a payoff of D + P(1 - Xp) - (3 * gk ), wherein P is defined as the selling price of the personal assets that are not exempted.

This study advances the natural person default literature in two important ways. First, it examines the relationship between changing home equity values and applying for the WSNP program. Second, this study also focuses on the relationship between opportunistic behavior and the Dutch insolvency law. The contribution to understand insolvency abuse is in investigating whether or not households exhaust other options prior to the WSNP. Based on the simple models mentioned in this section, it is expected that the willingness of households to cooperate in the amicable process is the deciding factor. Furthermore an increased significance is expected of having a mortgage with negative equity on failing the amicable process, due to the crisis. The stigma effect was unable on being measured and is therefore left out of the regression. The researcher was unable on getting into direct contact with the participants to quantify this effect.

3. Methodology & Data

3.1 Methodology Description

This paper combines several statistical and econometrical techniques to enable interpretation of the model, by testing the null hypothesis with the use of a probit model. This model is estimated with the use of the Maximum Likelihood Estimator (hereafter MLE). If the model is correctly specified MLE is consistent, asymptotically normal, and asymptotically efficient. The reason why this model is preferred originates from the dependent variable, which is binary. The model is constructed as follows:

Pr(Y=1|Cooperative, Mortgage, X)= 𝜙𝜙(𝛽𝛽0+ 𝛽𝛽1𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖,𝑡𝑡 (3.1)

+ 𝛽𝛽2𝑀𝑀𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀𝐶𝐶𝑀𝑀𝐶𝐶𝑖𝑖,𝑡𝑡+ 𝛽𝛽3𝑋𝑋𝑖𝑖,𝑡𝑡)

FailAmProc is an indicator of whether household i failed the amicable process (=1) or not (=0) in year t. Failing the amicable process is a function of cooperative behavior of the debtor, which is assessed by a curator according to his own professional judgment, whether the household has a mortgage, and the vector X. It should be mentioned that judging a participant as uncooperative is not an arbitrage process. The debtor is allowed to appeal against this decision in court.

Where X is a vector of input variables that operationalize resource input and household structure, characterized by gender, being non-white, head age, education higher than MBO, marital status, household size, being employed, entrepreneur, mental limitations, uncollateralized debt, foreclosure debt and negative equity. Table 2 provides a closer description about the metric unit and information of these variables.

Table 2 Selected Variables

Name Variable Metric Description

FailamProc Yes = 1 Fail of the amicable process

Cooperative Yes = 1 Whether participant was obstructing the process Mortgage Yes = 1 Whether the participant has a mortgage Vector X contains the following control variables

Negative equity Yes = 1 Presence of negative equity

UncollDebt Amount in Euro Uncollateralized Debt without Mortgage Foreclosure Debt Amount in Euro Debt due to foreclosure

Family Size Number of Persons People in households

Head Age Years Age of head household

Married Yes = 1 Marital Status

Employed Yes = 1 Job

Entrepreneur Yes = 1 Own business owner

Mental Limitations Yes = 1 Mentally Ill, receiving social welfare Non-white Yes = 1 All other races other than native Dutch Educ>Mbo Yes = 1 Education higher than MBO

Male Yes = 1 Gender

Number of Observations = 120 household years

3.2 Data Description

Sentinel Zorg, a receiver (Dutch: bewindvoerder) that negotiates between debtors and creditors, provided data for this study. They are located in Amsterdam, and the same applies for their clients who reside around the same area. In order to conduct further analysis, I requested Sentinel Zorg whether their database could provide us two samples; one in 2008 and one in 2014. The former is considered as not affected by the financial crisis, while the latter observation still experiences the aftereffect of the crisis. The data for each year is randomly and independently selected from the available database. If it became apparent that a randomly selected household contains missing data, the selected household was removed from the analytic sample because missing values are undesirable while using the probit model (a total of 8 households were removed due to missing key data.) After sample construction was completed, the analytic sample included 60 individual households for which full information for both years was available for analyses. Summing up to 120 household-year observations

Variable selection and interpretation was constructed such to mitigate imperfect multicollinearity. Thus a clear demarcation was made between mortgages and uncollateralized debt, and between debt that originates from foreclosure and uncollateralized debt. Another aspect of interest in the process of variable selection concerns the efficiency the model; adding redundant variables gives inefficient parameters. Nevertheless, the presence of biasness is undesirable in the model, therefore the selected variables are regressed in three different models.

3.3 Hypotheses

In line with the above outlined methodology description, the two hypotheses that will be tested in this paper focuses on the relationship between negative home equity and failing the amicable process. As such, the null and alternative hypotheses are formulated as follows:

𝐻𝐻0: 𝐹𝐹𝐶𝐶𝐶𝐶𝐹𝐹 𝐶𝐶𝑜𝑜 𝐶𝐶ℎ𝐶𝐶 𝐶𝐶𝑎𝑎𝐶𝐶𝑎𝑎𝐶𝐶𝑎𝑎𝐹𝐹𝐶𝐶 𝐶𝐶𝐶𝐶𝐶𝐶𝑎𝑎𝐶𝐶𝑝𝑝𝑝𝑝 𝑑𝑑𝐶𝐶𝑑𝑑 𝑛𝑛𝐶𝐶𝐶𝐶 𝑎𝑎ℎ𝐶𝐶𝑛𝑛𝑀𝑀𝐶𝐶

𝐻𝐻1: 𝐹𝐹𝐶𝐶𝐶𝐶𝐹𝐹 𝐶𝐶𝑜𝑜 𝐶𝐶ℎ𝐶𝐶 𝐶𝐶𝑎𝑎𝐶𝐶𝑎𝑎𝐶𝐶𝑎𝑎𝐹𝐹𝐶𝐶 𝐶𝐶𝐶𝐶𝐶𝐶𝑎𝑎𝐶𝐶𝑝𝑝𝑝𝑝 𝐶𝐶𝑝𝑝 ℎ𝐶𝐶𝑀𝑀ℎ𝐶𝐶𝐶𝐶 𝐶𝐶𝑛𝑛 2014

3.4 Hypotheses testing

Testing whether a structural change has taken place in failing the amicable process post the crisis is done with the likelihood ratio test (hereafter LRT). By comparing the log likelihood functions of the unrestricted model (ln 𝐿𝐿𝐿𝐿𝑈𝑈) and that

of the restricted model (ln 𝐿𝐿𝐿𝐿𝑅𝑅). The unrestricted probit model contains both the

2008 and 2014 observations, while the restricted model considers both samples separately. This proposition is captured in the following formula:

𝐿𝐿𝐿𝐿 = −2 ln�(𝐿𝐿𝐿𝐿𝑈𝑈𝑡𝑡=1+ 𝐿𝐿𝐿𝐿𝑈𝑈𝑡𝑡=2 ) − 𝐿𝐿𝐿𝐿𝑅𝑅𝑡𝑡=1+2�~𝜒𝜒2(𝐾𝐾)

= −2 ln(𝐿𝐿𝐿𝐿𝑈𝑈− 𝐿𝐿𝐿𝐿𝑅𝑅)

LR = −2 ln 𝜆𝜆 (3.2)

Where 𝜆𝜆 =𝐿𝐿𝑅𝑅𝑈𝑈

𝐿𝐿𝑅𝑅𝑅𝑅 𝑤𝑤𝐶𝐶𝐶𝐶ℎ 0 ≤ 𝜆𝜆 ≤ 1, and the ln is the natural logarithm The test statistic

restrictions. The null hypothesis sets all coefficients, except that of the intercept, equal to zero.

3.5 Interpretation of the coefficients

The model is a standard normal cumulative distributed function (hereafter c.d.f.). The stigma in model 3.1 symbolizes the c.d.f. of a standard normal random variable with error terms that are independent and normally distributed. The probit coefficients are the change in the z value associated with a unit change in the regressors. Although the effect of regressors on the z value is linear, its effect on the probability is nonlinear since they appear inside the c.d.f., because of that they cannot be estimated with ordinary least squares. Nevertheless, the assumptions made earlier on MLE allows constructing the (asymptotic) t tests and confidence intervals.

The marginal effect of the coefficient on the dependent variable depends on the type of variable. For continuous variables the average marginal effect (AME) for homoscedasticity is;

𝐴𝐴𝑀𝑀𝐴𝐴 = _𝑁𝑁1 ∑𝑁𝑁𝑖𝑖=1𝜙𝜙(𝑋𝑋𝑖𝑖𝛽𝛽)𝛽𝛽 (3.3)

Thus an infinitesimal change of a continuous variable changes the probability that FailAmProc takes the value of one by X percentage points. And for dummy variables; 𝐴𝐴𝑀𝑀𝐴𝐴 = 1 𝑁𝑁 �[𝜙𝜙�𝑋𝑋𝑖𝑖𝛽𝛽|𝑋𝑋𝑖𝑖𝑘𝑘 = 1� − 𝜙𝜙�𝑋𝑋𝑖𝑖𝛽𝛽�𝑋𝑋𝑖𝑖𝑘𝑘 = 0�] 𝑁𝑁 𝑖𝑖=1 (3.4)

A change of the dummy variable from zero to one changes the probability that FailAmProc takes the value of one by X percentage points.

4. Results & Analysis

4.1 Descriptive Statistics

Below one can find the descriptive statistics of the variables included in the probit estimation, for both the years 2008 and 2014. Households averaged 2.5 persons and the mean age of the sample was 45 years old across observations. Debt due to foreclosure and uncollateralized debt consisted of an average of €8,114.00 and €16,187.86 respectively. 38 percent of the sample possesses a mortgage and 25 percent of those mortgages were facing negative equity. Women, non-whites and persons with a Mbo degree or lower exhibit a higher representation in the sample. The high presence of applicants who have some kind of mental limitation in the sample can be explained due to other services that Sentinel Zorg provides, i.e. personal counseling and nursing.

Table 3 Descriptive Statistics Unrestricted Model

Variable Metric Mean Std. Dev Min Max

FailamProc _{Yes = 1 0.500 0.502 0 1}

Cooperative Yes = 1 0.758 0.430 0 1

Mortgage Yes = 1 0.375 0.486 0 1

Negative equity Yes = 1 0.250 0.435 0 1

UncollDebt Amount in € 16,188 24,524 0 149,372 Foreclosure Debt Amount in € 7,843 18,993 0 102,206

Family Size Number 2.508 1.506 1 8

Head Age Years 44.700 15.538 21 76

Married Yes = 1 0.508 0.502 0 1

Employed Yes = 1 0.467 0.501 0 1

Entrepreneur Yes = 1 0.200 0.402 0 1

Mental Limitations Yes = 1 0.292 0.456 0 1

Non-white Yes = 1 0.617 0.488 0 1

Educ>Mbo Yes = 1 0.392 0.490 0 1

Male Yes = 1 0.450 0.499 0 1

Table 4 exhibits the mean differences across the two years. On average, failing the amicable process increased in 2014. These findings are partly aligned with the literature review. We notice an increase in both foreclosure and uncollateralized debt, more younger households experience financial difficulties. Non-whites are in both samples a majority and we witness, on average, in 2014, an increase of ten percent. And are women more represented in the latest sample. However, other variables do not correspond with the findings in the literature review. Both the variables mortgage and negative equity show minimum differences across both samples, while it was expected that the latest sample would show an increase. The same holds true for unemployment, no differences are witnessed on average.

Table 4 Descriptive Statistics Restricted Model

2008

2014

Variable Metric Mean Dev Std. Mean Std. Dev

FailamProc _{Yes = 1 0.467 0.503 0.533 0.503} Cooperative Yes = 1 0.733 0.446 0.783 0.415 Mortgage Yes = 1 0.350 0.481 0.400 0.494 Negative equity Yes = 1 0.217 0.415 0.283 0.454 UncollDebt Amount in € 13,957 19,069 18,418 28,971 Foreclosure

Debt Amount in € 5,915 13,761 9,869 23,076 Family Size Number 2.850 1.494 2.167 1.452 Head Age Years 50.250 15.534 39.150 13.781

Married Yes = 1 0.583 0.497 0.433 0.500 Employed Yes = 1 0.483 0.504 0.450 0.502 Entrepreneur Yes = 1 0.267 0.446 0.133 0.341 Mental Limitations Yes = 1 0.333 0.475 0.250 0.437 Non-white Yes = 1 0.567 0.499 0.667 0.475 Educ>Mbo Yes = 1 0.333 0.475 0.450 0.502 Male Yes = 1 0.583 0.497 0.317 0.469

Number of Observations = 60 household years

Despite that the outcome of some elementary variables are not as how theory predicted, it would premature to draw any conclusion. Further statistical analysis is required in order to find structural changes.

4.2 Unrestricted Probit Regression

Table 5 reports the results of the estimated probit regression, calculated with the software program Stata. The columns distinguish three independent regressions which all feature mortgage and cooperative as main variables, and different intermixtures of control variables. Each incorporated variable has a coefficient, a standard error between brackets, and a significance notation ranging between one and ten percentage points.

Table 5 Probit Regression of all observations (2008 and 2014) with Fail Amicable

Process as Dependent Binary Variable

Unrestricted Model (1) (2) (3) Cooperative -1.501 (0.353)*** -2.169 (0.433)*** _{-2.112 (0.487)***} Mortgage 0.769 (0.267)*** -0.443 (0.460) -0.442 (0.487) Negative equity 0.999 (0.537)** _{1.015 (0.569)*} UncollDebt 0.000 (0.000)*** 0.000 (0.000)** Foreclosure Debt 0.000 (0.000)*** 0.000 (0.000)** Family Size 0.152 (0.125) Head Age -0.001 (0.011) Married -0.173 (0.416) Employed 0.355 (0.366) Entrepreneur 0.291 (0.568) Mental Limitation 0.159 (1.058) Non-white -0.895 (0.391) Educ>Mbo -0.101 (0.371) Male -0.036 (0.386) Intercept 0.918 (.340)*** 0.632 (0.402) 0.171 (0.840) Pseudo R2 _{0.213 0.528 0.550} Log Likelihood -65.49 -39.25 -37.39 N 120 120 120 *** p= <0.01; **p=<0.05, *p<0.10

The robustness of the variable cooperative becomes evident in table 5. In all three regressions it is observe that if participants are willing to provide all information to the curator, and don’t sabotage the amicable process, chances of failing the amicable process decreases significantly in each model. In the case of mortgages the opposite appears to be true; neither robustness nor significance are to be found in the other two models. Thus having a mortgage is not necessary inherent to failing the amicable process. However, in juxtaposition with having

negative equity on the mortgage, it becomes significant in model 2 and 3 with five and ten percent respectively. The first regression is considered as incomplete due to a limited choice of regressors, the goodness of fit measured with the Pseudo R2 _{exhibits a low value of 21 percent points. Despite that a high Pseudo R}2 value does not necessarily indicate a good fit, it, however, measures for proximity of the model to the observed data. An increase of additional regressors is coherent with an increase of the Pseudo R2_{, as model two and three display.} Having uncollateralized or foreclosure debt are both considered to be robust and significant. However, the extent of that is questionable. The small level of observations might be the cause of that: Out of the 120 observations, 25 participants encountered negative equity. The same holds for the remaining control variables, they’re not significantly different from zero. Substantially large standard errors are not found, as is expected after computing the correlation table. A printout can be found in the appendix. There seems to be no reason to assume that variable interactions suffer from imperfect multicollinearity due to the relatively low sample correlations.

A more thorough analysis of the effect of the regressors on failing the amicable process is conducted by computing the marginal effects in table 6. Setting all of the regressors to their mean values, the predicted probability of failing the amicable process (=1) becomes equivalent to 65 percentage points. Estimating that same probability but now considering the case in where mortgage and negative equity are both set at one, and the remaining regressors to their mean values, indicates a predicted probability of 81 percent. This sample test endorses the importance of having a mortgage with negative equity as a main reason for failing the amicable process.

Table 6 Probability FailAmProc equals 1

Variable

Mean

Value

Variable

Mean

Value

Cooperative

_0.758

Cooperative

_0.758

Mortgage

_0.375

Mortgage

₁

Negative equity

_0.250

Negative equity

₁

UncollDebt

_16,187

UncollDebt

_16,187

Foreclosure Debt

_7,843

Foreclosure Debt

_7,843

Family Size

_2.508

Family Size

_2.508

Head Age

_44.7

Head Age

_44.7

Married

_0.508

Married

_0.508

Employed

_0.467

Employed

_0.467

Entrepreneur

_0.200

Entrepreneur

_0.200

Mental

Limitations

0.292

Mental Limitations

0.292

Non-white

_0.617

Non-white

_0.617

Educ>Mbo

_0.392

Educ>Mbo

_0.392

Male

_0.450

Male

_0.450

Margin

Delta-Method

Std. Err.

Margin

Delta-Method

Std. Err.

Pr(FailAmProc=1)=

0.652 (0.078)*** Pr(FailAmProc=1)=

0.809 (0.117)***

*** p= <0.01; **p=<0.05, *p<0.10

As is expected after the probit regression, the marginal effects presented in table 7 exhibit the same significance level of the regressors. Being cooperative has a substantial effect on succeeding the amicable process, if the dummy variable changes from zero to one, the probability of succeeding rises by 78 percentage points. That same effect is witnessed with negative equity, where the probability of failing the amicable prices rises by 38 percent points in the presence of negative equity. The remaining continuous and dummy variables display little till no significance level.

Table 7 Marginal Effects on Failing Amicable Process

Variables

Marginal Effects

Cooperative

_{-0.780 ***}

Mortgage

_-0.163

Negative equity

_{0.375 *}

UncollDebt

_{0.000 ***}

Foreclosure Debt

_{0.000 ***}

Family Size

_0.056

Head Age

_0.000

Married

_-0.064

Employed

_0.131

Entrepreneur

_0.107

Mental Limitations

_-0.059

Non-white

_-0.039

Educ>Mbo

_-0.037

Male

_0.013

Observations

120

*** p= <0.01; **p=<0.05, *p<0.10

4.3 Restricted Model

Table 8 presents the restricted probit regression estimated separately acros 2008 and 2014. As is expected, the robustness of the variable cooperative becomes evident both in significance and negative effect, prior and post to the financial crisis. Possessing a mortgage is, along with the unrestricted model, considered as not significant from zero in both samples. Nevertheless, if the variable negative equity is included in the model, the distinction between both samples becomes notable. According to the provided data, negative equity could be considered as insignificant prior to the crisis. Whereas in 2014 it becomes evident that owning a mortgage with negative equity is considered as significant from zero at five and ten percent level.

The sole control variable that is considered significant in the 2008 model is whether a participant is an entrepreneur. Being an entrepreneur and experiencing a downturn contributes to a higher probability of failing the amicable process. This peculiar observation cannot be derived from the 2014 sample: the variables that contribute to a higher chance of failing the amicable

process are all debt related. Which corresponds to the notion that in an advert event, such as an economic downturn, the debt level is preeminent in quantifying vulnerability. However, a remark should be included that despite that uncollateralized and foreclosure debt are considered as significant from zero, their contribution to failing the amicable process are considered as negligible. Which does not corresponds with the literature review stated above. Debt burden is considered as a primary cause in filling for bankruptcy. This phenomenon might be caused due to the small amount of observation.

Table 8 Restricted Probit Regression with Fail Amicable Process as Dependent Binary Variable 2008 2014 (1) (2) (3) (1) (2) (3) Cooperative -2.19 (0.59)*** -2.56 (0.69)*** -3.41 (1.82)* -1.05 (0.47)** -1.53 (0.71)** -2.60 (1.45)*** Mortgage -1.31 (0.43)*** 0.29 (0.91) 2.59 (2.57) 0.42 (0.35) -2.15 (1.55) -5.98 (4.39) Negative equity 0.04 (1.23) -1.45 (1.29) 3.06 (1.45)** 7.20 (4.57)* UncollDebt 0.00 (0.00)*** 0.00 (0.00) 0.00 (0.00)** 0.00 (0.00)** Foreclosure Debt 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)*** 0.00 (0.00)** Family Size 0.49 (0.59) 0.65 (0.54) Head Age 0.11 (0.10) -0.04 (0.05) Married -0.34 (1.65) -0.24 (1.33) Employed 1.99 (2.25) -0.22 (0.98) Entrepreneur 2.76 (1.06)* -2.89 (2.07) Mental Limitations -6.00 (4.23) 2.07 (1.06) Non-white -1.40 (1.10) -0.89 (1.44) Educ>Mbo -0.24 (0.78) -2.25 (1.59) Male -1.05 (1.47) 3.06 (1.52) Intercept 1.19 (.54)** 0.07 (0.67) -9.71 (9.05) 0.78 (0.47)* 0.37 (0.66) -1.66 (3.51) Pseudo R2 0.388 0.7678 0.619 0.10 0.52 0.76 Log Likelihood -25.43 -9.57 -6.70 -37.18 -19.8 -10.01 N 60 59 60 60 60 60 *** p= <0.01 **p=<0.05 *p<.10

4.4 Likelihood Ratio Test

The complete models (3) in table 5 and 8 also incorporate the log likelihood figures that are necessary for the LRT. The LRT is used to test whether sufficient statistical evidence is found to reject the null hypothesis whether failing the amicable process has increased. Once rejected, it can be concluded that failing the amicable process has increased after the financial crisis. Disregarding the intercept, the model includes 14 variables that are added as restrictions. A test will be conducted with a Chi-squared distribution that adheres to the following critical values: 21.06, 23.69 and 29.14 to the significance level of respectively ten, five and one percentage points.

As can be observed, the unrestricted LR is given as -37.49 and the restricted LR’s are -6.7 and -10.1, respectively for 2008 and 2014. What follows from equation 3.1 is 𝜆𝜆 = 0.4469. Taking the ln of this figure (natural logarithm) and applying the remaining of equation 3.1, gives an outcome of 1.611. This value is considered in the critical region, therefore we cannot reject the null hypothesis. Given the data, it appears that there is no significance to be found at ten percent. Despite support of the theory, the power of the probit regression is considered as not sufficient. This can be explained due to the low amount of observations.

5. Discussion and Conclusion

The purpose of this study was to investigate whether the ongoing decline in housing prices contributes to opportunistic behavior of households, in the case of negative equity on their mortgages. Furthermore, this paper examined the role of cooperation in the amicable process, and how the Dutch insolvency law diminishes incentives to cooperate with the creditors. The appointed factors that are recognized as detrimental for the incentives are the exemptions, total amount of debt, level of garnishment and future income. It is considered that households make rational decisions, given the available information, cognitive ability of the participant and time. Realizing that creditors have little till no negotiation power in the amicable process, thus creating a moral hazard risk where debtors are willing to sabotage the amicable process is reinforced in the empirical results. The level of significance and robustness of cooperating is considered as the main cause in succeeding the amicable process. The justification for this rigorous policy towards creditors is to increase the force creditors to be cooperative during the amicable process. Once this fails, they risk losing much more of the original principle once the debtor enters the WSNP program.

Strategic defaulting on a negative equity mortgage implies that the default process is not only a matter of inability to pay, but rather of economic decision making, i.e. opportunity cost of renting, present value of the mortgage, and whether the future human capital exceeds the total amount of debt. The empirical results obtained in this study indicate causality between failing the amicable process and having negative equity, this support however is considered as weak and marginal. However, the theory is able on explaining this causality and it can be expected that if the power of the empirical tests increases, the significance becomes more evident. From the theory it becomes apparent that negative equity is necessary but not a sufficient condition in strategic defaulting. The data exhibits that most observations that failed the amicable process with a mortgage, also experience negative equity. The overall conclusion is that the existence of the three-year WSNP program is one of the reasons that contribute to the decrease of the amicable process success rate. Policy recommendation would be to make the amicable process less dependable on the willingness to cooperate of the debtor.

6. Bibliography

Berkowitz, J., & Hynes, R. (1999). Bankruptcy Exemptions and the Market for Mortgage Loans*. The Journal of Law and Economics, 42(2), 809-830. Calem, P. S., & Mester, L. J. (1995). Consumer behavior and the stickiness of

credit-card interest rates. The American Economic Review, 1327-1336. Dietz, F. J. (1998). Economie en beleid. Boom Koninklijke Uitgevers.

Domowitz, I., & Sartain, R. L. (1999). Determinants of the consumer bankruptcy decision. The Journal of Finance, 54(1), 403-420.

Fay, S., Hurst, E., & White, M. J. (2002). The household bankruptcy decision. American Economic Review, 706-718.

Fay, S., Hurst, E., & White, M. J. (2002). The household bankruptcy decision. American Economic Review, 706-718.

Foote, C. L., Gerardi, K., & Willen, P. S. (2008). Negative equity and foreclosure: Theory and evidence. Journal of Urban Economics, 64(2), 234-245.

Guiso, L., Sapienza, P., & Zingales, L. (2009). Moral and social constraints to strategic default on mortgages (No. w15145). National Bureau of Economic Research.

Godwin, D. D. (1999). Predictors of households’ debt repayment difficulties. Financial Counseling and Planning, 10(1), 67-78.

Gross, D. B., & Souleles, N. S. (2002). An empirical analysis of personal

bankruptcy and delinquency. Review of Financial Studies, 15(1), 319-347. Han, S., & Li, G. (2011). Household borrowing after personal bankruptcy. Journal

of Money, Credit and Banking, 43(2‐3), 491-517.

Hynes, R. M. (2004). Why (consumer) bankruptcy. Ala. L. Rev., 56, 121. Jungmann, N. (2006). De Wsnp: bedoelde en onbedoelde effecten op het

Livshits, I., MacGee, J., & Tertilt, M. (2007). Consumer bankruptcy: A fresh start. The American Economic Review, 402-418.

Markell, B. A. (2001). Sorting and Sifting Fact From Fiction: Empirical Research and The Face of Bankruptcy: The Fragile Middle Class: Americans in Debt By Teresa A. Sullivan, Elizabeth Warren and Jay Lawrence Westbrook. Am. Bankr. LJ, 75, 145-447.

Simon, H. A. (1957). Models of man; social and rational.

Warren, E. (2003). Growing Threat to Middle Class Families, The. Brook. L. Rev., 69, 401.

Warren, E., Thorne, D., & Woolhandler, S. (2005). Illness and injury as

contributors to bankruptcy. Cambridge, MA: Harvard Medical School. Warren, E. (2003). Growing Threat to Middle Class Families, The. Brook.

L. Rev., 69, 401.

White, M. J. (2007). Abuse or Protection-Economics of a Bankruptcy Reform under BAPCPA. U. Ill. L. Rev., 275.

Ziegel, J. S. (2001). A Canadian Perspective The Fragile Middle Class: Americans in Debt. By Teresa A. Sullivan, Elizabeth Warren. Tex. L. Rev., 79, 1241- 2177.

Zhu, N. (2011). Household consumption and personal bankruptcy. The Journal of Legal Studies, 40(1), 1-37.

7. Appendices

Descriptive Statistics 2008 and 2014

2008