PRELIMINARY AND INCOMPLETE: PLEASE DO NOT CITE WITHOUT PERMISSION

Transcription

1 PRELIMINARY AND INCOMPLETE: PLEASE DO NOT CITE WITHOUT PERMISSION The Financial Crisis at the Kitchen Table: Recent Trends in Household Debt and Credit By Meta Brown, Andrew Haughwout, Donghoon Lee and Wilbert van der Klaauw Federal Reserve Bank of New York 33 Liberty Street New York, NY 145 First Draft: September 23, 29 Abstract We describe a detailed credit-report dataset derived from a unique longitudinal quarterly panel of individuals and households from 1999 to 29. Our panel is based on a nationally representative 5% random sample of all individuals with credit reports (usually aged 19 and over). We also sampled all other individuals living at the same address as the primary sample members, allowing us to track household-level credit and debt for a random sample of US households that includes over 37 million individuals. Our findings indicate that after a large increase in total consumer debt after June 1999, total debt peaked at $12.5 trillion in 28Q2; the total debt balance has seen a small but persistent decline since then. This overall pattern corresponds to that for mortgages, credit cards and auto loans, but differs from those of student loans and balances on home equity lines of credit, which actually have continued to increase during the past year. The decline in the overall debt balance has been accompanied by a large recent drop of in the total number of open accounts. This drop is caused by a surprisingly large increase during the past year in account closings, especially credit card accounts, as well as a significant decline in account openings. We present evidence that the latter mainly reflects a decline in the demand for credit, while the former appears to have been initiated by banks. While overall debt has started to fall, however, we find that defaults and delinquencies have continued to increase with mortgage related delinquencies actually accelerating. As of June 3 29, approximately 12% ($1.5 trillion) of the total outstanding debt balance was delinquent (more than 3 days late), with 8% ($1 trillion) being seriously delinquent (at least 9 days late). While the decline in overall debt and continued increase in delinquencies is common across states, the extent of both changes varies considerably, with the largest changes occurring in states that witnessed the largest housing booms and declines, including Arizona, California, Florida and Nevada. The views expressed are those of the authors and not of the Federal Reserve Bank of New York or of the Federal Reserve System. Sungsu Kim and Sergiu Laiu provided excellent research assistance.

2 Introduction Policy makers, the press and academic researchers have recently showed urgent interest in the liabilities side of household balance sheets. By most accounts, the ongoing financial crisis began in the residential mortgage market, as increasingly large numbers of borrowers, especially in the nonprime market segment, became delinquent on their mortgage payments. The increase in these delinquencies and the enormous rise in residential mortgage foreclosures soon developed into a full-blown financial crisis, and led to one of the sharpest contractions in US history. While many features of the financial system played a role in these developments, household behavior was clearly a fundamental contributor. In an effort to improve understanding of the characteristics and trends in household liabilities, the New York Fed has created a research dataset from consumer credit reports from Equifax, which is one of the three major credit bureaus in the US (the other two are Experion and TransUnion). This paper describes the data and the sample, provides some preliminary findings, and lays out an agenda for their future use. The FRBNY-Equifax Consumer Credit Panel The dataset we describe here, which we call FRBNY-Equifax Consumer Credit Panel, comprises a 5% random sample of US individuals with credit files and all of the family members of those 5%. In all, the data set includes files for more than 15% of the population, or approximately 37 million individuals. We observe information from the credit reports for these individuals each quarter for the last 1 years, with current data through June 29. The data will 2

3 continue to be updated every quarter in the future: data for 29Q3 will be available by October 31, 29. The sampling exploits randomness in the last two digits of individuals Social Security numbers. In each period, the primary sample consists of persons whose Social Security numbers match five of the 1 possible combinations of digits (, 1, 2,, 99). This procedure ensures that the panel is dynamically updated in each quarter to reflect new entrants into credit markets. In addition, the data provider matches the primary individual s mailing address to all records in the data in order to capture information about other members of the primary individual s household. These individuals are also added to the sample. This procedure enables us to track individuals and households consistently over time, thus allowing us to study richer dynamics of consumer debt and related policy issues at both the individual and household levels. Our credit report data includes residential location at the census block level and the individual s month and year of birth. The data contain detailed information of each individual mortgage loan, including Origination date Original balance Current balance Current (scheduled) payment Current status (i.e., current, 3 days delinquent, etc.) While the mortgage information in the dataset is very detailed and we believe, complementary to loan-level information available from sources like LoanPerformance and LPS (McDash), it differs in important ways from these. In particular, the mortgage information does 3

4 not indicate the seniority of individual mortgage loans. On the other hand, because the FRBNY- Equifax panel data are collected at the borrower level, they offer a different perspective on mortgage debt than is available in standard loan-level datasets. In addition to information on debts secured by residential real estate, the data set includes somewhat more aggregate data on individuals and households other loans, such as credit cards, auto loans and student loans. Here, the data include the following: Total number of each kind of account (e.g., the total number of bank-issued credit cards) The credit limit on each type of account (e.g., the combined credit limit on all credit cards) 1 Total balance on each type in each status (e.g., the total student loan balance that is current, 3 days delinquent, etc.) More general information on the credit report includes the following: Indicators for whether the individual has a foreclosure or bankruptcy within the last 24 months and ever on the report An indicator for whether the individual has any accounts in collections and the amount of collection The Equifax credit score (analogous to the well known FICO score) 1 This field is known as the high credit amount in the credit report data. It refers to either the credit limit (for credit cards, home equity lines of credit and other revolving debt) or the highest balance (for mortgages, auto loans and other installment debt). 4

5 Initial Results The initial findings of the household debt situation in the US can be summarized in two ways. First, since the second quarter of 28, the US households have gone through a continuing deleveraging process, resulting in a decrease in the aggregate consumer debt balance, which now (as of June 3, 29) stands at about $12 trillion. Secondly, delinquency and defaults continue to increase and accelerate, resulting in about 12% of the total debts being delinquent now compared to about 8% one year ago, and 4% in 25. The increase in delinquency and defaults are most apparent in housing related loans, and also in the states that experienced rapid housing booms. Figure 1 shows the total debt reported on the credit reports for the last 1 years by various types of loans. Total household debt decreased by about 2 to 3 percent over the last year, and the current total debt is $12.2 trillion. Mortgage related debts account for nearly 8% of the total debt, with the rest being composed of credit cards, auto loans and student loans. It is interesting to see that, despite the general decrease in total debt, student loans and home equity revolving debt, also known as home equity lines of credit (HELOCs), actually increased a little bit compared to the same quarter of last year. Even though the deleveraging is a nationwide phenomenon, there are significant variations across states. Figure A1 shows that the increase and subsequent decrease in household debt is most apparent in states that had a housing market boom such as California, Nevada, Florida and Arizona. Figure A3 shows the differences in June 29 household debt composition over states. Many interesting differences emerge, particularly the relatively high share of housing-related debt in boom states like Arizona and California, and the noticeably low share in Texas, where HELOCs are not allowed. 5

6 The deleveraging of household debt is also shown in the number of open accounts. Figure 2 shows the decrease in the number of various open accounts. This overall decline was caused by a decrease in the number of new accounts (blue line in Figure 3) and also, at the same time, closings of existing accounts by consumers and lenders (red line in Figure 3). The decline in Equifax inquiries (green line in Figure 3), which represent potential lenders credit record pulls following an application for credit, suggests that individuals are applying less frequently for credit, tentative evidence that a decline in demand for credit is partly responsible for net reductions in open accounts. Figure 4 shows the decrease in credit limit on credit card accounts and home equity revolving accounts, which is likely a result of lender actions. In the credit card case, total credit limit decreased by 15 to 2 percent over the last year, pushing up the utilization rate (balance divided by credit limit) by about 2 percent. On the other hand, it is the increase in the balance that raised the utilization rate on the home equity revolving accounts: credit lines are essentially unchanged Along with the decrease in household debt, delinquency and defaults increased rapidly with little sign of stabilization through June, as shown in Figure 5. In 25, delinquent balance accounted for only 4% of the total balance, with serious delinquency, defined by 9 days late or more accounting for only 2% of the total balance. However, these figures tripled and quadrupled respectively, accounting for 12% and 8% of the total balance as of the most recent quarter. It's interesting to see the deterioration of household debt started as early as 26, and has accelerated since then. Delinquency and defaults in mortgage loans basically drive the patterns in total debt in Figure 5, given that mortgage debts account for three quarters of the total debt. Taking a look at different loan types, Figure 6 shows that the deterioration of the debts is a common theme across all types of debts, but it's quite interesting that the speed of deterioration 6

7 is most noticeable in the mortgage debts and home equity revolving debts. Looking across the different states, we again confirm that the four states of California, Nevada, Florida and Arizona, riddled with housing problems, stand out against the rest of the states in Figure A2. The deterioration of mortgage debts and other household debts is naturally reflected in an increase in foreclosures and personal bankruptcies. Figure 7 shows the quarterly number of new foreclosures and bankruptcies nationwide, where the annual rate of foreclosures is now over 2 million people, which is about 1% of the total population with credit reports. (It is important to reiterate that this measure of new foreclosures is at the individual level. It is the number of individuals with a foreclosure newly added to their credit report, as opposed to the number of mortgages or houses with a foreclosure notice, a more commonly reported figure.) Note the spike in the personal bankruptcy rate in 25, which is due to the change in the bankruptcy law that made filing for bankruptcy more difficult after that year. Figure 8 reports quantiles of the Equifax credit score distribution over the last ten years. It is difficult to see much movement in any but tenth percentile. A zoom-in of that quantile level is shown in figure A4 for the US and various states. Here a general pattern of decrease from 1999 to 23, then increase to 25, then decrease is discernible, but Texas is again an outlier in terms of both level and time pattern. Conclusion and Directions for Future Research The FRBNY-Equifax Consumer Credit Panel allows analyses of individual and household behavior that have heretofore been difficult to conduct. The rich geographic detail allows analysts to examine spatial patterns in the data in ways that have not been previously 7

8 possible. We intend to produce aggregate reports for the US and each individual Federal Reserve District, and distribute these at a quarterly frequency. In addition, we hope to collaborate in the production of maps that will provide a convenient visual presentation of some of the key measures. The data also allow analysis of many pressing questions in consumer and household finance. We can use the data to provide insight into the effects of a number of current policy initiatives, including new regulations on credit card accounts bankruptcy reform mortgage loan modifications mortgage refinancing incentives An additional question is the long-term individual and household consequences of a foreclosure or bankruptcy. While these phenomena have become much more prevalent in recent years, we observe them throughout the 1 year history of the data, and can examine their effects on individual outcomes like future access to and terms of credit. 8

9 Data Dictionary The FRBNY-Equifax Consumer Credit Panel consists of detailed credit-report data for a unique longitudinal quarterly panel of individuals and households from 1999 to 29. The panel is a nationally representative 5% random sample of all individuals with a social security number and a credit report (usually aged 19 and over). We also sampled all other individuals living at the same address as the primary sample members, allowing us to track household-level credit and debt for a random sample of US households. The resulting database includes approximately 4 million individuals in each quarter. More details regarding the sample design can be found in Lee and van der Klaauw (29). A comprehensive overview of the specific content of consumer credit reports is provided by Avery, Calem and Canner (Federal Reserve Bulletin, February 23). The credit report data in our panel primarily includes information on accounts that have been reported on by the creditor within 3 months of the date that the credit records were drawn each quarter. Thus accounts that are not currently reported on are excluded. Such accounts may be closed accounts with zero balances, dormant or inactive accounts with no balance, or accounts that when last reported had a positive balance. The latter accounts include accounts that were either subsequently sold, transferred, or paid off as well as accounts, particularly derogatory accounts, that are still outstanding but on which the lender has ceased reporting. According to Avery et al, the latter group of noncurrently reporting accounts with positive balances when last reported accounted for approximately 8% of all credit accounts in their sample, and for the vast majority of these accounts, and particularly mortgage and installment loans, additional analysis suggested they had been closed (with zero balance) or transferred. 1 Our exclusion of the latter accounts is comparable to some stale account rules used by credit reporting companies, which treat noncurrently reporting revolving and nonrevolving accounts with positive balances as closed and with zero balance. All figures shown in the tables and graphs are based on the 5% random sample of individuals. To reduce processing costs, we drew a 2% random subsample of these individuals, meaning that the results presented here are for a.1% random sample of individuals with credit reports, or approximately 3, individuals. 2 In computing several of these statistics, account was taken of the joint or individual nature of various loan accounts. For example to minimize biases due to double counting, in computing individuallevel total balances, 5% of the balance associated with each joint account was attributed to that individual. Per-capita figures are computed by dividing totals for our sample by the total number of people in our sample, so these figures apply to the population of individuals who have a credit report. 1 Avery et al (23) found that for many nonreported mortgage accounts a new mortgage account appeared around the time the account stopped being reported, suggesting a refinance or that the servicing was sold. Most revolving and open non-revolving accounts with a positive balance require monthly payments if they remain open, suggesting the accounts had been closed. Noncurrently reporting derogatory accounts can remain unchanged and not requiring updating for a long time when the borrower has stopped paying and the creditor may have stopped trying to collect on the account. Avery et al report that some of these accounts appeared to have been paid off. 2 Due to relatively low occurrence rates we used the full 5% sample for the computation of new foreclosure and bankruptcy rates.

10 Loan types. In our analysis we distinguish between the following types of accounts: mortgage accounts, home equity revolving accounts, auto loans, bank card accounts, student loans and other loan accounts. Mortgage accounts include all mortgage installment loans, including first mortgages and home equity installment loans (HEL), both of which are closed-end loans. Home Equity Revolving accounts (aka Home Equity Line of Credit or HELOC), unlike home equity installment loans, are home equity loans with a revolving line of credit where the borrower can choose when and how often to borrow up to an updated credit limit. Auto Loans are loans taken out to purchase a car, including Auto Bank loans provided by banking institutions (banks, credit unions, savings and loan associations), and Auto Finance loans, provided by automobile dealers and automobile financing companies. Bankcard accounts (or credit card accounts) are revolving accounts for banks, bankcard companies, national credit card companies, credit unions and savings & loan associations. Student Loans include loans to finance educational expenses provided by banks, credit unions and other financial institutions as well as federal and state governments. 3 The Other category includes Consumer Finance (sales financing, personal loans) and Retail (clothing, grocery, department stores, home furnishings, gas etc) loans. Our analysis excludes authorized user trades, disputed trades, lost/stolen trades, medical trades, child/family support trades, commercial trades and, as discussed above, inactive trades (accounts not reported on within the last 3 months). Total debt balance. Total balance across all accounts, excluding those in bankruptcy. Number of open, new and closed accounts. Total number of open accounts, number of accounts opened within the last 12 months. Number of closed accounts is defined as the difference between the number of open accounts 12 months ago plus the number of accounts opened within the last 12 months, minus the total number of open accounts at the current date. Inquiries. Number of credit-related consumer-initiated inquiries reported to Equifax in past 6 months. Only hard pulls are included, which are voluntary inquiries generated when a consumer authorizes lenders to request a copy of their credit report. It excludes inquiries made by creditors about existing accounts (for example to determine whether they want to send the customer pre-approved credit applications or to verify the accuracy of customer-provided information) and inquiries made by consumers themselves. Within each industry of auto finance, mortgage, and utilities (excluding wireless), multiple inquiries in 3-day periods count as one inquiry. Note that not all inquiries are reported to all three credit reporting companies 4, and the reporting practices among credit reporting companies may have changed during the period of analysis. High Credit and Balance for Credit Cards. Total amount of high credit on all credit cards held by the consumer. High credit is either the credit limit, or highest balance ever reported during history of this 3 The student loan delinquency rates shown in Figure 6 reveal a more volatile pattern and an overall higher delinquency rate prior to 23, which may reflect a change in reporting behavior where lenders previously may not have reported on loans on which repayment may have been deferred for a period of time (see Avery et al, 23). 4 Typically, each inquiry is reported to only one of the three credit reporting companies only, with different creditors reporting to different credit reporting companies.

11 loan. As reported by Avery et al (23) the use of the highest-balance measure for credit limits on accounts in which limits are not reported likely understates the actual credit limits available on those accounts. High Credit and Balance for HE Revolving. Same as for credit cards, but now applied to HELOCs. Credit utilization rates (for revolving accounts). Computed as proportion of available credit in use (outstanding balance divided by credit limit), and for reasons discussed above are likely to overestimate actual credit utilization. Delinquency status. Varies between current (paid as agreed), 3-day late (between 31 and 6 day late; not more than 2 payments past due), 6-day late (between 61 and 9 days late; not more than 3 payments past due), 9-day late (between 91 and 12 days late; not more than 4 payments past due), 12-day late (at least 121 days past due; 5 or more payments past due) or collections, and severely derogatory (any of the previous states combined with reports of a repossession, charge off to bad debt or foreclosure). Not all creditors provide updated information on payment status, especially after accounts have been derogatory for a longer period of time. Thus the payment performance profiles obtained from our data may to some extent reflect reporting practices of creditors. Percent of balance 9+ day late. Percent of balance that is either 9-day late, 12-day late or severely derogatory. New Foreclosures. Number of individuals with foreclosures first appearing on their credit report during the past 3 months. Based on foreclosure information provided by lenders (account level foreclosure information) as well as through public records. Note that since borrowers may have multiple real estate loans, this measure is conceptually different from foreclosure rates often reported in the press. For example, a borrower with a mortgage currently in foreclosure would not be counted here if he receives a foreclosure notice on an additional mortgage account. In the case of joint mortgages, both borrowers reports indicate the presence of a foreclosure notice in the last 3 months, and both are counted here. New Bankruptcies. New bankruptcies first reported during the past 3 months. Based on bankruptcy information provided by lenders (account level bankruptcy information) as well as through public records. Collections. Number and amount of 3 rd party collections (i.e. collections not being handled by original creditor) on file within the last 12 months. Includes both public record and account level 3 rd party collections information. As reported by Avery et al (23), only a small proportion of collections are related to credit accounts with the majority of collection actions being associated with medical bills and utility bills. Equifax Risk Score. Credit score computed by Equifax s credit scoring model. The Equifax Risk Score, like the FICO score, ranges from 3-85, with a higher score being viewed as a better risk than someone with a lower score.

12 References Avery, R.B., P.S. Calem, G.B. Canner and R.W. Bostic, An Overview of Consumer Data and Credit Reporting, Federal Reserve Bulletin, Feb. 23, pp Lee, D. and W. van der Klaauw, An introduction to the FRBNY-Equifax Consumer Credit Panel, [in progress].

13 Fig. 1: Total Debt Balance and its Composition Trillions of Dollars Trillions of Dollars Mortgage HE Revolving Auto Loan Credit Card Student Loan Other (3%).5 (4%).8 (7%).7 (6%).7 (6%) 9.1 (74%) (9%) (1%) (8%) (69%) :Q1 :Q1 1:Q1 2:Q1 3:Q1 4:Q1 5:Q1 6:Q1 7:Q1 8:Q1 9:Q1

14 Fig. 2: Number of Accounts by Loan Type Millions 25 Millions 5 2 Credit Card (right axis) Mortgage (left axis) Student Loan (left axis) Auto Loan (left axis) HE Revolving (left axis) 99:Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1 3

15 Fig.3: Total Number of New and Closed Accounts and Equifax Inquiries Millions 2 Millions 2 Number of Accounts Closed within 6 Months Number of Inquiries within 3 Months Number of Accounts Opened within 6 Months 5 99:Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1

16 Fig. 4: Credit Limit and Balance for Credit Cards and HE Revolving Trillions of Dollars 4 CC Limit CC Balance HELOC Limit HELOC Balance Trillions of Dollars 4 23 % 3 27 % 3 Utilization Rate % 5 % 54 % % 99:Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1

17 Fig. 5: Total Balance by Delinquency Status Percent 1% Severely Derogatory 12-day late 9-day late 6-day late 3-day late Current Percent 1% 95% 95% 9% 9% 85% 85% 8% 8% 75% 99:Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1 75%

18 Fig. 6: Percent of Balance 9+ Days Delinquent (by Loan Type) Percent 15 Percent 15 1 Credit Card 1 5 Student Loan Mortgage 5 Auto Loan HE Revolving 99:Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1

19 Fig.7: Number of New Foreclosures and Bankruptcies Thousands 1,25 Foreclosures Bankruptcies Thousands 1,25 1, 1, :Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1

20 Fig. 8: US Credit Score Distribution Score 85 9% Quantile Score % Quantile % Quantile % Quantile % Quantile :Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1 5

21 Fig. A1: Total Debt Balance by State (Per Capita*) Thousands of Dollars National Average NJ IL Thousands of Dollars 1 NV CA 75 AZ FL 5 MI NY 25 OH TX PA 25 99:Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1 *Based on the population with a credit report

22 Fig. A2: Balance 9+ Days Delinquent by State Percent 2 Percent 2 15 NV FL 15 1 National Average AZ 1 5 NY OH MI TX IL PA CA NJ 99:Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1 *Based on the population with a credit report 5

23 Fig. A3: Debt Composition by State (29:Q2) Percent 1% 9% 8% 7% 6% 5% 4% 3% 2% 1% Mortgage HE Revolving Auto Loan Credit Card Student Loan Other Percent 1% 9% 8% 7% 6% 5% 4% 3% 2% 1% % US AZ CA FL IL MI NJ NV NY OH PA TX %

24 Fig. A4: 1% Quantile Credit Score by State Score 57 Score AZ MI IL PA CA NJ NY FL National Average OH NV 99:Q1 1:Q1 3:Q1 5:Q1 7:Q1 9:Q1 *Based on the population with a credit report TX 51 49