Beruflich Dokumente
Kultur Dokumente
Macro Picture
Ritika Sinha
1 Introduction
Historically, long run data on US GDP shows alternating patterns of in‡ation and unemployment as asso-
ciated with business cycles. Delinquency rates on consumer credit card loans are seen to follow the same
pattern as national unemployment rates but with lower variations (see Graph-1). Ausubel writes about
the cyclical nature of credit card defaults in tandem with GDP levels, working at a macro-economic level.
The motivation of this paper is to test if the movement of unemployment rates and other macroeconomic
2 Literature Review
Sissoko (2006) uses a generic growth model to evaluate macroeconomic determinants of delinquency rates
over a ten year period from 1994 to 2004. The paper …nds that unemployment, interest rates, debt to personal
income ratio, delinquency rate lagged one period are positively and signi…cant related to delinquency rates,
while credit supply is positively but not signi…cantly related. Cointegration tests were performed on the
dependent and independent variables and a long-term equilibrium relationship was found. GDP growth and
policy decisions may a¤ect delinquency through interest rates and unemployment rates (both statistically
signi…cant).
A 2003 paper by Agarwal and Liu examines panel data for 700 thousand accounts, using variables
representing individual account characteristics like balance, credit limit, interest rate, state-speci…c “shock”
variables like divorce rates, bankruptcy …lings, and health coverage, legal variables and county-speci…c unem-
ployment rates. It uses the proportional hazard model to estimate the conditional probability of delinquency
at time t and concludes that unemployment rates are signi…cant in determining delinquency rates in all
speci…cations, irrespective of the lag length. It also concludes that the line of credit as a proxy for the supply
1
side is negatively related to delinquency.
Grieb, Hegji and Jones (2001) use the …ndings of Paquin and Weiss (1998) on bankruptcy as a determinant
of credit card delinquency as a starting point. The paper uses the …rst di¤erences of the dependent and
independent variables (both relative and absolute) to correct for the presence of autocorrelation and the
results vary. It tests the idea of selective default when borrowers tend to default earlier on credit card loans
to ful…ll other debt obligations and …nd that the e¤ect tends to be felt after a lag of one period of increase
in revolving debt.
There is a multitude of papers on the e¤ect of bankruptcy …ling decision on credit card delinquency.
Stavins uses logit regression on SCF 2004 panel data to assess both the demand (borrower’s characteristics)
and supply (lender) sides of credit card markets, including the e¤ect of recent bankruptcy legislation. Ausubel
writes about the self-correcting nature of credit card lending, making bankruptcy legislation insigni…cant in
default rate determination. Both Ausubel and Stavins explain the adverse selection problem faced by banks
in determining default risk of borrowers. Black and Morgan talk about the changing characteristics of credit
3 Model Description
Credit card delinquency is postulated to be dependent on several macro economic variables, supply side
factors and personal reasons. Several papers evaluate the signi…cance of personal factors like divorces, bank-
ruptcy, health insurance costs, age, etc. using cross-section individual survey data (Agarwal and Liu). This
model uses aggregate time series data to evaluate the e¤ect of economy-wide factors like the unemployment
rate, all credit card interest rates, debt-income ratio (DSR), bankruptcies; and the supply-side e¤ect through
where:
2
delrate is Delinquency rate on credit card loans; All commercial banks SA1
intrate is Commercial Bank Interest Rate on Credit Card Plans - All Accounts NSA
t is Time trend
The Board of Governors’delinquency rate is computed by taking the dollar amount of credit card loans 30
days or more past due (accruing and not accruing interest) for the quarter and dividing by the level of credit
card loans outstanding at the end of the quarter. The unemployment rate (in percentage terms) refers to the
proportion of the civilian labor force that is unemployed. The relationship being tested for signi…cance here
is: as unemployment rises, more people would tend to default on their credit card payments. The household
debt service ratio (DSR) is the ratio of debt payments to disposable personal income. Debt payments consist
of the estimated required payments on outstanding mortgage and consumer debt. The variable for interest
rate refers to the average interest rates charged by commercial banks on all credit card loans. These interest
rates are the terms of credit set by the credit card issuers (supply side), as a fee for credit card borrowing.
Di¤erential interest rates are often charged, depending on the credit risk that the borrower poses. Revolving
Credit outstanding refers to the amount of revolving credit lent out by commercial banks. This is used as
a proxy for the supply side of credit card loans. Log of Bankruptcy is the natural log of bankruptcy …lings
The model was estimated using quarterly data for all variables over an 18 year period, from 1991(Q1) to
2008(Q2). The results from the OLS regression show a goodness of …t of 54.91% (R2 = 0.5491, adjusted R2
= 0.5062). All of the variables seem to be positively signi…cant in determining delinquency rates on credit
cards, except lrevcred, which is signi…cant but negatively related, all evaluated at the 5% level of signi…cance.
1 SA means seasonally adjusted
3
The data shows strong positive serial correlation (lag 6), as proven by the Breusch-Godfrey test LM
2
statistic = 31.120, against the critical value = 12.59. Also, when the residuals (uhat) are regressed
against the lagged residuals (luhat), luhat is signi…cant with the t statistic = 6.56. To correct for serial
correlation, the Newey-West robust standard errors were used with a lag of 6 periods, su¢ cient for quarterly
data. Interestingly enough, as the bandwidth changes (6, 20, 55), the signi…cance of the variables may also
change thus showing that the standard errors are not robust for this model. Adjusting for this shift using
the Kiefer-Vogelsang tables at 97.5% where 2 b = 0.1, the following results are found for lag (6).
Unemployment rates are statistically signi…cant (with a t-statistic of 2.64) and positively related ( 1 =
0.49) to credit card delinquency rates. A one percentage point increase in unemployment would bring about
a 0.49 percentage point increase in delinquency. The intuition is that due to unemployment, credit card loans
become an alternate source of funds for consumer spending (Agarwal and Liu). Borrowers tend to increase
credit card spending to tide over temporary loss of income, with the intention of repaying the debt as soon
as they are employed again. If unemployment persists beyond one period, unemployed borrowers are likely
to be delinquent on their credit card interest payments. In addition to this, the demographic of workers
who use credit card is changing from only white-collared workers to a larger number of blue-collared workers
(Black and Morgan). As these sections are more prone to and more a¤ected by lay-o¤s and phases of job-loss,
unemployment now has a greater impact on credit card delinquency than earlier. According to Sissoko, the
economic climate in the aggregate economy and policy decisions for …scal and monetary expansion or control
The debt-service ratio (DSR) has a high coe¢ cient ( 2 = 42.69) but is not statistically signi…cant. The
ratio of debt payments to personal disposable income depends on federal tax rates, marginal propensity to
consume and earnings (Sissoko). One would suppose that the increasing proportion of debt to disposable
personal income would force individuals to default on credit card interest payments and other debt payments
as well. However, both, Grieb, Hegji and Jones as well as Agarwal and Liu …nd that consumers tend to
selectively default on credit card payments to tide over other debt obligations like mortgage payments.
Unemployment is also related to the notion of selective default as it causes a loss of earnings and income
4
Interest rates, as expected, are extremely signi…cant in determining delinquency rates, at all lag lengths.
3 = 0.57, thus indicating that a one percentage point increase in interest rates on credit cards will increase
the delinquency rate by 0.57 percentage points. As interest rates increase, credit card spending will become
more expensive, thus causing more rate-sensitive people to default on credit card loans to channelize the
income towards repayment of other debt. Where interest rates are concerned, banks often face an adverse
selection problem. According to Stavins, as interest rates increase from a lower level, consumers who orig-
inally did not want to borrow would exit the market, due to the high cost of loans. On the other hand,
risky borrowers with a higher probability of default would enter the market, raising delinquency rates. The
process would be cyclical if banks chose to charge higher interest rates to account for these default losses. As
adverse selection takes place, banks choose not to alter the interest rates too much, either way (Ausubel).
Log Revolving Credit (supply of loans) has a negative relationship ( 4 = -4.16) but is insigni…cant in
determination of credit card delinquency rates. The relationship shows that a one unit increase in revolving
credit outstanding would decrease delinquency rates by 4.16%. This relationship is contrary to Sissoko’s
…ndings but similar to Agarwal and Liu’s results on line of credit. This relationship is possibly because
consumers now have a larger amount of loans forthcoming and thus would not need to default on loans
to channel that income elsewhere. Also, market e¢ ciency would dictate that as supply of loans increases,
Log Bankruptcy is another variable that is positively signi…cant in explaining delinquency rates. A one
unit increase in bankruptcy …lings leads to a 0.8% rise in default rates ( 5 = 0.801) According to Ausubel,
banks have begun to realize the importance of default risk that arises on lending to borrowers with high
debt-income ratios since they are more likely to …le for bankruptcy. Delinquency rates were at their highest
during the bankruptcy crisis in the new millennium. However, banks continue to lend to risky borrowers at
as long as the interest rate spread is high enough to make it a pro…table venture. When they become too
risky, banks are unwilling to lend to ‘subprime’borrowers, which in turn will motivate fewer people to …le
5
5 Conclusion
The model shows that macroeconomic factors and the economic environment have an impact on credit card
delinquency rates. Unemployment rates, interest rates and bankruptcy …lings are all positively signi…cant;
DSR is positively related but insigni…cant and revolving credit (supply side) is negatively related and insignif-
icant in explaining delinquency rates. In a period of GDP growth, when the Monetary and Fiscal policies
of the government try to control the in‡ationary trend, unemployment rates rise and so do bankruptcy and
delinquency rates. On the other hand, when the economy is in a recession and policy tries to inject growth
and con…dence into the economy by increasing employment, delinquency rates will fall. Thus delinquency
rates are in‡uenced by macroeconomic factors, although the e¤ect of these may be lagged depending on how
6
References
Agarwal, Sumit and Liu, Chunlin, ‘Determinants of Credit Card Delinquency and Bankruptcy:
Ausubel, Lawrence M., ‘Credit Card Defaults, Credit Card Pro…ts, and Bankruptcy’, American
Ausubel, Lawrence M., ‘Adverse Selection in the Credit Card Market’, Working Paper,
Ausubel, Lawrence M., ‘Personal Bankruptcies Begin Sharp Decline: Millennium Data Update,’
Black, Sandra E. and Morgan, Donald P., ‘Meet the new borrowers’, Current Issues in
Economics and Finance, Federal Reserve Bank of New York, February 1999
Grieb, Terrance, Hegji, Charles, Jones, Steven T., ‘Macroeconomic factors, consumer behavior, and
Kiefer, Nikolas and Vogelsang, Timothy, ‘A New Asymptotic Theory for Heteroskedasticity-Autocorrelation
Sissoko, Macki, ‘Determinants of Delinquency Rate on Commercial Banks Mortgage and Credit Card Debts
Stavins, Joanna, 2000. ‘Credit card borrowing, Delinquency, and Personal bankruptcy’, New England
Economic Review, Federal Reserve Bank of Boston, issue July, pp. 15-30
7
Tables
OLS Coef OLS SE NW Lag 6 NW Lag 20 NW Lag 55
Variable Name