1. The Impact of Credit Availability and Interest Rate on Consumer Car Expenditures:
Evidence from the Consumer Expenditure Survey
Zachary Chen
University of California, Santa Barbara
Department of Economics
March 2015
Acknowledgements This paper is written as the undergraduate senior thesis at the University of
California, Santa Barbara. I am grateful for this opportunity given by the Department of
Economics. I would like to thank my thesis advisors Shelly Lundberg and Paulina Oliva for their
valuable comments on this research. I am also indebted to Kelvin Yang from the College of
Engineering for his assistance in dealing with the data.
2. Abstract
This paper analyzes the impact of credit availability and interest rate on total consumer
car expenditures. I study such effects using comprehensive consumption data from the Consumer
Expenditure Survey. With bank lending and the fed funds rate as credit measurements, I estimate
the effect of credit constraints on consumer car expenditures and purchases. Consistent with
other empirical findings, I conclude that credit availability has positive effects on car
expenditures. Furthermore, I also study the effect of consumer demographics and show that
income and family size have positive associations with car expenditure and purchase decisions.
However, unlike previous findings on housing expenditures, I find no evidence that the effect of
credit availability is stronger in the last decade than in the 1990s. In addition, I do not find
consistent evidence on the relationship between interest rate and car expenditures. I suggest that
these findings may be the results of low positive expenditures between 2000 and 2013. More
research is still needed to determine the reasons behind such declines in car consumption.
3. Introduction and Literature Review
Throughout the years, economists have utilized various economic models to study the
characteristics of consumer durable goods. Unlike nondurable goods, consumers continue to
receive a flow of benefits over many years after the purchase of durables. During economic
booms and recessions, one should expect durable expenditures to be procyclical with economic
fluctuations. In fact, the relative importance of consumer durables decreases by about 20 percent
from economic boom to recession (Reed and Crawford, 2014). In addition, the purchases of
durable goods such as new cars are usually infrequent. Consumers typically postpone their
purchases of new automobiles when income fluctuates frequently, which shows consumers’
desires to smooth durable consumptions over a longer period (Hassler 2001). In economic
theory, households are assumed to maximize expected lifetime utility of their desired stock of
durables, subject to current and expected lifetime income. In the absence of credit constraints,
households could smooth the consumption of durables over lifetime; such rationale could be
explained by the permanent income hypothesis proposed by Milton Friedman in 1957. In
contrast to nondurable goods, expected increase in future income is followed by the decline in
current durable expenditures, because consumers plan to subsequently increase durable
consumption in the future and to smooth durable consumptions overtime (Cah, Ramey, and Starr
1995). These findings support the well-known evidence that consumer durables are highly
sensitive to income changes.
However, besides changes in income, the presence of credit constraints could possibly
hinder consumers’ objective to effectively smooth their consumption of durable goods. In fact,
economic research often cites that the presence of credit constraints serve as the contradiction to
the permanent income hypothesis. Credit constraints could alter consumers’ optimal borrowing
decisions and prevent households from fully optimizing their intertemporal allocation of durable
4. goods overtime (Deaton 1992, Attanasio, Goldberg, and Kyriazidou 2008). Previous academic
research has mainly focused on the role of credit conditions on housing expenditures, which is
one of the most important durable goods in the economy. One common justification for the boom
in housing demand during the early 2000s is the easy availability of credit in the form of low
interest rates. (Himmelberg, Mayer and Sinai 2005, Taylor 2007, and Taylor 2009). Such high
credit availability and low interest rate environment during the early 21st
century initiated many
research discussions on whether credit expansions have lead to structural changes in the housing
market. Some researchers argue that the increase in credit availability in the credit market could
be accounted for most of the boom in housing expenditures (Shiller 2005, Favilukis, Ludvigson,
and Van Nieuwerburgh 2010). Other research show that there is no convincing evidence that the
fluctuation in credit conditions could explain the recent surge in housing expenditures (Glaeser,
Gottlieb, and Gyourko 2010). They claim that other things such as psychology, mortgage
securitization problems, and fraud incentives could most explain the surge in the housing market.
Since many research discussions are focused on the housing market, it is natural to
wonder whether credit constraints are binding on the consumption of other durable goods.
Besides the housing market, automobiles are one of the more important representations of
consumer durables. Similar to the housing market, the interest rate in the automobile market has
been decreasing over the last two decades (Figure 1). In fact, the average quarterly finance rate
on new automobiles decreased from 9.3% between 1990 and 1999 to 6.9% between 2000 and
20131
. If credit expansion and interest rates had tremendous impact on durables as housing, it is
reasonable to suspect that they might have the similar effects on car expenditures.
1
This is the “Finance Rate on Consumer Installment Loans at Commercial Banks, New Autos 48 Month Loan”. The
data comes from the Federal Reserve Economic Data.
5. Many previous academic literatures have studied the relationship between credit
conditions and car expenditures. Older studies such as (Witt and Johnson 1986, Attanasio,
Meghir, and Weber1997) find that high-purchase credit restrictions have a significant negative
impact on new car expenditures. Using maturity, loan size, and interest rate as credit
measurements, Attanasio, Goldberg and Kyriazidou (2008) conclude that credit conditions do
affect the overall purchases of new cars. They also find that lower income households are
responsive to credit availability but not to interest rates. However, these authors find no evidence
that younger households are more credit constrained than older ones. Interest rate should also
negatively affect household’s consumption in that higher interest rate makes current
consumption more expensive relative to future consumption (Friedman, 1957). However,
Walden (2013) finds no significant relationship between interest rate and consumer vehicle
expenditures. The current consensus among economic literature is that the increase in credit
availability has positive effects on overall consumer car expenditures; however, there are no
consistent findings in the effects of interest rate.
The purpose of this paper is to examine whether the increase in credit availability and
low interest rate have similar effects on consumer car expenditures as in housing expenditures.
Using comprehensive detailed data from the Consumer Expenditure Survey (CE), Federal
Reserve Economic Data (FRED), and the Senior Loan Officer Opinion Survey on Bank Lending
Practices, I test the hypothesis that credit constraints have stronger effects on car expenditures
during the 2000s than the 1990s. I plan to test such hypothesis by comparing the estimates
between these two decades.
Consistent with other studies, I find that credit availability has positive effects on
consumer car expenditure and purchase decisions. Similar to other durables, I also conclude that
6. car expenditures are highly sensitive to income changes. While finding negative interest rate
effects from 1990-1999, I find no significant interest rate effects on car expenditures during the
2000s. Furthermore, although there is no apparent evidence that the relaxation of credit have
stronger effects on car expenditures over the last decade, this paper still contributes to the
ongoing economic discussions on the role of credit availability and interest rate in determining
consumer car expenditure and purchase decisions.
By comparing expenditure data from 1990-1999 and 2000-2013, I will first analyze the
impact of credit availability on consumer car expenditures from 2000-2013. I then estimate the
same regression models on household consumption data from 1990-1999. I further examine the
different effects between credit availability, real interest rates and car expenditures. Finally, I
conclude by addressing the current downward trend on car expenditures among younger
consumer households. I believe this is the one of the first economic papers to estimate and
compare the effects of credit availability and interest rates on consumer car expenditures
between the 1990s and the 2000s.
The remaining of the paper is organized into the following five sections. The first section
describes the empirical econometric models. The next part describes the data I am going to use in
this paper. The third section presents the descriptive statistics on the important variables in my
regression models. The fourth and fifth section of this paper will then present the empirical
results and conclusion.
7. Empirical Methodology
I use both tobit and logit regression models to test the hypothesis that credit condition and
interest rate have significant effects on car expenditures and purchase decisions. The tobit model
determines the total effect of credit lending conditions on consumer car expenditures. I choose
the tobit model because car expenditures at left-censored at zero. The logit model estimates the
change in the likelihood that consumer household makes a purchase. The dependent variable is
the real total car expenditures for the individual household from that particular quarter. In the
logit model, total car expenditures takes on the value of one if the consumers bought a car for the
quarter and zero if they did not.
To examine the effect of credit conditions on consumer car expenditures, most research
studies chose the real interest rate and one or two other credit attributes as a proxy for credit
conditions in the credit market (Grieves 1983). In this paper, I use two independent credit
variables as the determinants of consumer car expenditures. First, I use the net percentage change
in bank’s willingness to increase consumer loans and the real fed funds rate as proxies for credit
conditions in the economy. The bank’s willingness to increase lending measures the exogenous
change in the turn-down rate that consumers face at any particular quarter. That is, banks are
affecting the difficulties of access to consumer credit for any given quarter. This credit
measurement (Figure 2) comes from the Senior Loan Officer Opinion Survey on Banking
Lending Practices, which is often used as an assessment for credit availability in the economy
(Schref and Owen 1991, Lown and Morgan 2006, Bassett, Chosak, Driscoll, and Zakrajsek,
2013). Positive percentage change represents consumers are obtaining easy credit access
compared to the previous quarter, and negative percentage illustrates the opposite effect.
8. Therefore, I could utilize this variation in bank lending to capture the effect of credit availability
on consumer car spending and purchasing decisions.
Second, I consider many different interest rates in the economy and choose to use the
effective fed funds rate, adjusted for inflation through the GDP deflator, as the real interest rate
measurement. It is reasonable to assume in economics that general automobile interest rate might
be endogenous with the random error term in that it is partly influenced by consumers’ decisions
to borrow. As the leading interest rate indicator in the credit market, however, the fed funds rate
could plausibly be exogenous to consumer households. It is determined and targeted by the Fed,
which in principle could be independent of the error term. Figure 3 displays the nominal
movement of the effective fed funds rate over the last two decades. Different from bank lending,
the real interest rate captures the general price of borrowing in the credit market for consumer
households. Low interest rate represents low costs of borrowing while high interest rate increases
the price of borrowing on vehicle purchases.
To control for income variations, I include the natural logarithm of the household’s after-
tax income. Since car expenditures could be correlated with the time of the year, I include a
dummy variable for each quarter of the year to control for seasonal effects. I then include the
quarterly unemployment rate to control for business cycle fluctuations and macroeconomic
conditions that could affect car expenditures and purchases. Finally, I control for consumer
demographics by including the age of head of the household, age squared, and a dummy variable
for each family size. With the corresponding independent variables, I yield the following
empirical model:
(1)
9. where Yi is the dependent variable that measures total car expenditure at any particular quarter.
In the logit model, the dependent variable car expenditure Yi takes the following form:
where represents either the percentage change in bank’s willingness to lend or the real
interest rate. measures the total effect of credit availability and interest rate on total car
expenditures. After substituting in (2) into equation (1), then estimates the probability that
consumer households make a purchase on automobile given a one percentage change in credit
availability and interest rate. That is, the change in probability that Log(Inci) is the
natural logarithm of the household’s real after-tax income, as measured by the GDP deflator.
Unemployi measures the unemployment rate for the quarter. X2i represents a vector of additional
demographic variables that could potentially influence car expenditures. Lastly, is the
unobservable errors in the regression models that describe everything else that is not included in
the regression models, such as households’ preferences and tastes for cars.
I expect to find a positive relationship between credit conditions and car expenditures and
a negative effect on interest rate. As credit lending and availability increase, consumers should
become more likely to purchase a car and increase their marginal spending on car expenditures.
Conversely, as interest rate increases in the economy, consumers should decrease their car
expenditures and postpone their purchase decisions when future consumption seems more
valuable. I also believe that car expenditures and car purchase decisions are positively correlated
with income and negative correlated with business cycles.
10. Data Description
For the data between 2000 and 2013, I use the public-use micro data of the Consumer
Expenditure Survey (CE) from the Bureau of Labor Statistics (BLS). However, the BLS does not
release available micro-data on the years before 2000. Therefore, I use the Consumer
Expenditure Survey Family-Level Extracts from the National Bureau of Economic Research
(NBER) data between 1990 and 1999. John Sabelhaus started the extracts of this survey and Ed
Harris from the Congressional Budget Office continued the process. The third and fourth quarter
expenditure data from 1995 were not included in this paper.2
In summary, the CE is an important
federal survey that provides comprehensive detailed consumer expenditures, income, and
demographic characteristics for researchers to study consumer spending trends and preferences.
The CE also has good reliability and consistency, because the CE division from the BLS
frequently conducts research in improving data quality and addressing problems such as
declining response rate (Garner, McClelland, and Passero 2009 and GoldbenBerg 2009).
Furthermore, after comparing different consumer expenditure surveys in 33 countries, McBride
and To (2013) show that the U.S. CE had a response rate of about 73 percent in 2010, which is
13 percent higher than the average response rate across other countries. However, one of the
well-known potential weaknesses of the survey is that some estimates of expenditures are still
biased due to measurement errors and non-response by the consumer unit (Garner, McClelland,
and Passero, 2009).
Despite these shortcomings, the CE is still the largest expenditure survey that records
information on the spending habits of American households, which makes it useful in analyzing
consumer expenditures across different decades. The CE is one of the most reputable
2
This is due to changes in the sampling frame of the survey. See http://www.nber.org/data/ces_cbo.html for
detailed information on the survey.
11. consumption surveys in the United States. The BLS not only uses the survey to estimate
consumption expenditures but also utilizes it to calculate the consumer price index (CPI). The
CE samples approximately 7000 representative U.S. civilian households per quarter during the
2000s and 4000 households during the 1990s. Among other interviewing questions, the
consumers are asked “What was the amount paid for the vehicle after rebate and discount? How
much did you receive in wages before taxes? How much federal income tax was deducted from
your pay?” I use information like these on the CE to examine different vehicle spending patterns
and trends. This paper uses and compares sample individual households from 1990-1999 and
2000-2013.
Results
I provide the empirical results in this section of the paper. Table 1 presents the descriptive
statistics on the dependent and independent variables. I include basic consumer expenditures,
income, and demographic characteristics that may affect car expenditures and purchases.
Expenditure characteristics contain the real total car spending during a quarter and the number of
owned vehicles. The rest of the table presents sample statistics on bank lending, interest rate,
after-tax income, unemployment rate, age of the reference person, and the family size of the
household. Panel A and B illustrate some notable differences in consumer characteristics and
demographics between the 1990s and the 2000s. While the number of vehicles owned reduces
modestly from 2.1 to about 1.9, the average car expenditure decreases from around $2500
between 1990 and 1999 to about $300 between 2000 and 2013. In fact, the fraction of sample
households with positive car purchases significantly decline from 27% in 1990-1990 to a mere
2.03% between 2000 and 2013. This is a very surprising finding in that car expenditure and
purchase seem to have decreased dramatically over the last two decades. I will discuss such
12. findings later in the paper. In addition, the price of borrowing in the credit market is much higher
in the 1990s than in the 2000s. Specifically, the average finance rate on new automobile
decreases from 9.3% to 7%, and the fed funds rate decreases from 7% to around 3.5%. Table 1
also shows some changes in consumer characteristics. Real after-tax income increases from
$40,000 to $50,000 while family size has decreased from about 3.1 to 2.54. These summary
statistics illustrate the various changes in consumer expenditures, income, and characteristics.
Therefore, perhaps credit conditions and interest rate affect car expenditures and purchases
differently across the two periods. I will investigate such differences in the remaining part of this
section.
Table 2 shows the effect of bank lending on total car expenditures and purchases for both
2000-2013 and 1990-1999. As expected, columns (1) and (3) both show that credit availability
has positive effects on consumer car expenditures between the two periods. When bank’s
willingness to lend increases by one percent, total car expenditure increases by 28 dollars in the
2000s and 67 dollars in the 1990s. This finding confirms that bank’s approval to credit access
has positive effects and plays an important role in influencing household car expenditures.
Similarly, columns (2) and (4) confirm that the likelihood of car purchase increases with bank
lending as well. Based on these two estimates, I find that the probability to purchase a car given a
change in credit availability is roughly the same for households in both periods. However, unlike
housing expenditures, there is no convincing evidence that the effects of credit availability on car
expenditures are stronger during the 2000s.
Table 3 estimates the effect of real interest rate on total car expenditures. For the 1990s,
columns (3) and (4) estimate that interest rate has strong negative effects on car expenditures but
not on purchases decisions. Although the interest rate effects are negative, columns (1) and (2)
13. show that there is no significant relationship between interest rate and car expenditure/purchase
decisions. However, I believe that this finding does not necessarily mean that interest rate is
unimportant to car expenditure and purchase decisions. It could be that the insignificant result
was due to high sampling variance. Furthermore, there are no consistent findings among
economic literature between interest rate and vehicle expenditures (see Attanasio, Goldberg, and
Kyriazidou 2008 and Walden 2013). Therefore, it is not surprising to find that interest rate has no
effects on car expenditures.
Not surprisingly, both model specifications show consistent estimates for income and
family size. As expected, household income has strong positive effects on car expenditures. In
fact, income has the strongest effects on car expenditures and purchase decisions. For every one-
percentage increase in income, average consumer car expenditures increase by about 3000
dollars (2009 dollars). Given by the odds ratio, the likelihood of purchasing a car is also similar
among both groups of consumers. Therefore, both tobit and logit models convey that income is
certainly an important determinant of consumer car expenditures. This is consistent with most
theoretical and empirical findings in economic literature.
Both models further show significant differences in car spending and purchasing between
large and small households. In fact, families with more than seven members spend almost twice
as much on cars than households with only two family members. One possible explanation for
this finding is that larger families are more likely to be constrained by vehicles as the general
method of transportation. These households have higher car expenditures compare to smaller
households. This effect is also shown in the logit model where larger families are about one and
a half times more likely to purchase a car in both decades.
14. However, the empirical findings also show some surprising and striking results on the
impact of age and time of the year. First, in both model specifications, age has significant impact
on car expenditures in the 1990s but not between 2000 and 2013. This surprise discovery may
coincide with other recent studies on car expenditures among younger consumers. As this paper
shows, car expenditures have declined significantly from 2000-2013. It could be that consumers
are no longer driving as much in the 21st
century compare to previous decades, especially among
younger cohorts (Davis, Dutzik and Baxandall 2012, 2013). The continuation of this trend might
explain the inconclusive evidence on the relationship between age and car expenditures. Second,
the regression estimates show that consumers have higher car expenditures in the second and
third quarter during the 2000s. This consumption pattern is different from expenditure trends
from the 1990s, where most spending occurs during the first quarter of the year. I examine these
apparent differences shown in the regression tables by further looking into the micro data in the
CE. Between 2000 and 2013, the mean total car expenditure in the second and third quarter is
about $317 and $255 in first quarter. In the 1990s, the mean car expenditure is $3417 during the
first quarter and $3363 among other quarters. Although not included in the summary statistics
table, this descriptive data result confirms the seasonal estimates in both model specifications. I
find that consumers have higher car expenditures in the third quarter in the 2000s, whereas their
consumption is higher in the first quarter during the 1990s. These empirical estimates and
descriptive statistics may suggest changes in consumer preferences toward the timing of car
expenditures
In summary, consistent with other findings on car expenditures, I find strong evidence
that both credit availability and household income positively affect consumer car expenditures. I
conclude that high credit growth is associated with higher car expenditure. However, I find no
15. convincing evidence that the relaxation of credit availability has stronger effects during the
2000s. Although both specifications show negative associations, interest rate only has strong
negative effects during the 1990s. I also obtain consistent findings in the effect of family size but
not in age. Unlike the consensus findings on housing expenditures, I did not find convining
evidence that the expansion in credit availability over the last decade lead to higher expenditure
on car consumption. The reason for these findings may due to low positive purchases during the
2000s. Some recent studies have studied this declining trend in car consumption among
consumers. Neiva and Gifford (2012) cannot conclude whether such low purchases on
automobiles are due to changes in the taste of driving among younger consumers or due to the
effect of the Great Recession. Therefore, future studies are still needed to uncover more detailed
relationships between car expenditures, consumer preferences, and credit conditions.
Conclusion
In this paper, I investigate the relationship between credit availability and interest rate on
total consumer car expenditures. I exploit two exogenous variations that characterize credit
conditions in the credit market: the percentage of bank’s willingness to lend and the fed funds
rate. Bank lending signifies the variations in credit availability and fed funds rate represents the
price of borrowing for consumers in the credit market. Sample statistics reveal some notable
differences between the two decades. Real interest rate decreases from an average of 7% in
1990-1999 to 2.5% between 2000 and 2013. Average total car expenditure and the fraction of
households with positive car purchase also decline dramatically through the two decades.
Consumer demographics such as income have increased while family size has decreased over the
years.
16. My results are consistent with consensus findings on consumer durables. I find that credit
availability has positive effects on car expenditure and purchase decisions. In addition, I find
persistent positive effects on income and family size. However, in contrast to the findings on
housing expenditures, I cannot conclude that credit availability has stronger effects on car
expenditures during the 2000s than in the 1990s. I also find no significant relationship between
interest rate, age, and car expenditures in the 2000s. I suggest that such findings are due to low
positive car purchase between 2000 and 2013, where some recent studies have suggested that
consumers are no longer driving as much. However, the studies are inconclusive on whether such
declines are due to changes in consumer preferences or other reasons. Such investigation is left
for future studies on car expenditure, which will remain as a relevant topic for years to come.
17. References
Attanasio, O., Banks, J., Meghir, C., & Weber, G. (1999). Humps and Bumps in Lifetime
Consumption. Journal of Business & Economic Statistics, 17(1), 22-22.
Attanasio, O., Goldberg, P., & Kyriazidou, E. (2008). Credit Constraints In The Market For
Consumer Durables: Evidence From Micro Data On Car Loans. International Economic Review,
49(2), 401-436.
Bassett, W., Chosak, M., Driscoll, J., & Zakrajšek, E. (2013). Changes in bank lending standards
and the macroeconomy. Journal of Monetary Economics.
Case, K., Quigley, J., & Shiller, R. (2005). Comparing Wealth Effects: The Stock Market versus
the Housing Market. Advances in Macroeconomics.
Chah, E., Ramey, V., & Starr, R. (1995). Liquidity Constraints and Intertemporal Consumer
Optimization: Theory and Evidence from Durable Goods. Journal of Money, Credit, and
Banking, 27(1), 272-287.
Davis, B., Dutzik, T., & Baxandall, P. (2012). Transportation and the New Generation Why
Young People Are Driving Less and What It Means for Transportation Policy.
Deaton, A. (1992). Saving and Income Smoothing in Cote d'Ivoire. Working Papers from
Priceton, Woodrow Wilson School - Development Studies.
Favilukis, J., Ludvigson, S., & Nieuwerburgh, S. (2010). THE MACROECONOMIC EFFECTS
OF HOUSING WEALTH, HOUSING FINANCE, AND LIMITED RISK-SHARING IN
GENERAL EQUILIBRIUM. NBER WORKING PAPER SERIES, Working Paper 15988.
Friedman, M. (1957). The Permanent Income Hypothesis. A Theory of the Consumption
Function.
Garner, T., McClelland, R., & Passero, W. (2009). Strengths and Weaknesses of the Consumer
Expenditure Survey from a BLS Perspective.
Glaeser, E., Gottlieb, J., & Gyourko, J. (n.d.). CAN CHEAP CREDIT EXPLAIN THE
HOUSING BOOM? NBER WORKING PAPER SERIES, Working Paper 16230.
Goldberg, K., & Ryan, J. (2009). June 29, 2009 Evolution and Change in the Consumer
Expenditure Surveys: Adapting Methodologies to Meet Changing Needs.
Grieves, R. (1983). The Demand for Consumer Durables. Journal of Money, Credit and
Banking, 15(3), 316-316.
Hassler, J. (2000). Uncertainty and the Timing of Automobile Purchases. Scandinavian Journal
of Economics.
18. Himmelberg, C., Mayer, C., & Sinai, T. (2005). Assessing High House Prices: Bubbles,
Fundamentals And Misperceptions. Journal of Economic Perspectives, 19(4), 67-92.
Lown, C., & Morgan, D. (2006). The Credit Cycle and the Business Cycle: New Findings Using
the Loan Officer Opinion Survey. Journal of Money, Credit, and Banking, 38(6), 1575-1597.
Neiva, R., & Gifford, J. (2012). DECLINING CAR USAGE AMONG YOUNGER AGE
COHORTS: AN EXPLORATORY ANALYSIS OF PRIVATE CAR AND CAR-RELATED
EXPENDITURES IN THE UNITED STATES.
Reed, S., & Crawford, M. (2014). How does consumer spending change during boom, recession,
and recovery? Beyond the Numbers, 3(15).
Schreft, S., & Owens, R. (1991). Survey Evidence of Tighter Credit Conditions: What Does It
Mean? Economic Review.
Taylor, J. (2007). HOUSING AND MONETARY POLICY. NBER WORKING PAPER SERIES,
Working Paper 13682.
Taylor, J. (2009). The Need to Return to a Monetary Framework. Business Economics, 44(2), 63-
72.
To, N., & McBride, B. (n.d.). A Comparison of Consumer Expenditure Survey. Federal
Committee on Statistical Methodology (FCSM) Research Conference.
Walden, M. (2013). Where did we indulge? Consumer spending during the asset boom. Monthly
Labor Review.
Witt, S., & Johnson, S. (1986). An econometric model of new-car demand in the UK.
Managerial and Decision Economics, 7(1), 19-23.
19. Appendix
Figure 1. The average finance rate on consumer installment loans, new 48 months auto
Figure 2. The percentage change in bank’s willinges to increase lending compare to last quarter
20. Figure 3. The average quarterly effective federal funds rate from 1990-2013
21. Table 1. Descriptive statistics
Mean (SD) Min Max
Panel A (2000-2013)
Real total car exp
Fraction of households
w/ positive purchase
293.62 (2656.79)
2.03%
0
-
120418.9
-
Credit (%)
Real interest rate
Average finance rate
on new auto
5.65 (13.82)
2.45 (2.42)
6.90 (1.30)
-47.2
0.07
4.13
28.8
7.94
9.64
Real after-tax income1
($10000)
5.0 (5.76) 0 142.14
Age in years 49.20 (17.48) 14 94
Family size
Number of vehicles
Unemployment rate
N = 400421
Panel B (1990-1999)
Real total car exp
Fraction of households
w/ positive purchase
Credit (%)
Real interest rate
Average finance rate
on new auto
Real after-tax income1
($10000)
Age in years
Family size2
Number of vehicles
Unemployment rate
(%)
N = 116931
2.54 (1.51)
1.86 (1.51)
6.32 (1.85)
3466.55 (8920.63)
26.8%
9.58 (10.29)
6.92 (2.13)
9.32 (1.22)
3.94 (4.36)
41.10 (18.74)
3.08 (1.69)
2.10 (1.69)
5.68 (1.09)
1
0
3.9
0
-
-14.8
4.12
7.54
0
14
1
0
4.1
30
24
9.9
184650
-
29.3
12.54
11.89
83.55
94
28
21
7.6
1
24704 observations with income < 0 are dropped
2
Family size is rounded to the nearest integer
Notes: Real variables are obtained by deflating nominal variables by the 2009 quarter 1 GDP
deflator. The results here are based on my own tabulations from the BLS CE public use micro-
data file (2000-2013) and the NBER CE Family-Level extracts files (1990-1999).
22. Table 2. Results on the effect of bank lending on total car expenditures
Dependent variable: real total car expenditures
2000 - 2013 1990 - 1999
Tobit
(1)
Logit (odds ratio)
(2)
Tobit
(3)
Logit (odds ratio)
(4)
Credit 27.69**
(13.69)
1.002**
(0.001)
66.96***
(10.16)
1.004***
(0.0008)
Real after-tax
Income (log)
3052.75***
(206.14)
1.184***
(0.015)
3730.28***
(116.99)
1.248***
(0.01)
Unemployment
Rate
-1061.61***
(104.48)
0.935***
(0.006)
-284.40***
(94.11)
1.026***
(0.007)
Family size
2
3
4
5
6
≥ 7
7912.688***
9866.76***
11672.65***
12319.04***
15184.94***
15071.5***
1.741***
2.003***
2.261***
2.365***
2.867***
2.916***
8413.16***
11931.05***
13364***
13459.03***
13138.48***
14234.27***
1.894***
2.608***
2.980***
3.132***
3.199***
3.383***
Age -119.02
(76.18)
0.991*
(0.005)
130.51***
(28.97)
1.007***
(0.002)
Age squared -0.16
(0.79)
1.000
(0.0005)
-2.47***
(0.33)
1.000***
(0.00002)
Quarter
2
3
4
2105.90***
2802.95***
715.58
1.144***
1.187***
1.039
-1380.89***
-542.09***
-708.48***
0.900***
0.949***
0.946***
N = 342658 342658 113495 113495
Notes: *** indicates statistically significant at the 1 percent level, ** at the 5 percent level, and *
at the 10 percent level. All regressions include a constant term and heteroskedastic robust
standard errors are in parentheses.
23. Table 3. Results on the effect of real interest rate on total car expenditures
Dependent variable: real total car expenditures
2000 - 2013 1990 - 1999
Tobit
(1)
Logit (odds ratio)
(2)
Tobit
(3)
Logit (odds ratio)
(4)
Real interest rate -173.18
(137.70)
0.991
(0.009)
-158.62***
(44.96)
0.998
(0.004)
Real after-tax
Income (log)
3046.56***
(206.34)
1.184***
(0.015)
3729.62***
(117.03)
1.249***
(0.01)
Unemployment
Rate
-1202.65***
(186.28)
0.928***
(0.006)
-163.41*
(92.01)
1.038***
(0.007)
Family size
2
3
4
5
6
≥ 7
7910.38***
9875.99***
11681.33***
12328.95***
15194.74***
15084.94***
1.741***
2.004***
2.261***
2.365***
2.867***
2.918***
8412.79***
11940.18***
13367.6***
13480.61***
13136.06***
14236.48***
1.893***
2.606***
2.977***
3.132***
3.377***
3.377***
Age -119.29
(76.18)
0.991*
(0.005)
131.13***
(28.98)
1.007***
(0.002)
Age squared -0.16
(0.79)
1.000
(0.0001)
-2.47***
(0.334)
1.000***
(0.00002)
Quarter
2
3
4
2193.28***
2862.11***
654.75
1.151***
1.192***
1.035
-999.17***
-352.15
-623.24**
0.9222***
0.961*
0.955**
N = 342658 342658 113495 113495
Notes: *** indicates statistically significant at the 1 percent level, ** at the 5 percent level, and *
at the 10 percent level. All regressions include a constant term and heteroskedastic robust
standard errors are in parentheses.