Abstract This paper offers a conceptual and partly empirical decomposition of the trendsin U.S. consumer expenditures on five communications and nine transportation subcategoriesbetween 1984 and 2002. We find that inflation nearly always increases unit prices. Incomeeffects are positive for all categories, meaning that these are all normal goods, not inferiorones. We speculate that taste changes have contributed to increasing expenditures in mostcategories, with the exception of out-of-town lodging, the public transit component of thepublic transportation category, and the old communication media categories of postage andreading. We suggest that production and technological changes have led to decreased unitprices in most categories. In the private vehicle operations categories, technological improve-ments dominate, so that expenditure shares have been decreasing despite increasing demand.Conversely, in the new media categories, taste changes dominate, so that expenditure sharesare increasing despite technological improvements which lower prices. The decompositionexplored here enhances our understanding of these important expenditure categories, andprovides a useful methodology with which to examine trends in other categories as well.
P. L. Mokhtarian (B)Institute of Transportation Studies, University of California, Davis,One Shields Avenue, Davis, CA 95616, USAe-mail: [email protected]
T. Lee, S. ChooThe Korea Transport Institute, 1160 Simindaero, Ilsanseo-gu, Goyang-si,Gyeonggi-do, 411-701, Korea
In policymaking, in scholarly research, and in everyday life, a great deal of attention isfocused on communications and transportation activities. Together they (broadly defined)consume around a quarter of personal income, and make an incalculable contribution toeconomic prosperity and individual and social well-being. On the other hand, both are-nas—but especially transportation—impose environmental costs on society, and these inducemuch of the attention paid to them. The communications arena is in the midst of whatcan reasonably be called a revolution in products and services, which prompts still morenotice. The transportation sector, too, has undergone substantial innovation over the pastfew decades, even if those changes are somewhat less “disruptive” than those in communi-cations. Furthermore, these two sectors are inextricably linked: both involve physical net-works that serve as conduits for human interaction, and as such enjoy a complex relationshipin which each can both substitute for and complement the other (Mokhtarian 1990, 2003,2009).
This relationship, too, has been the source of much attention, and motivates the pres-ent investigation. This paper grew out of a larger study whose purpose was to examine therelationships between consumer expenditures on communications and transportation, usingthe Almost Ideal Demand System (AIDS) of modeling consumer demands. Key modelingresults from that study appear in Choo et al. (2007a,b), and additional details relevant tothose papers and the present one can be found in Choo et al. (2006). This paper supportsthe other two by offering a more detailed analysis of the trends in consumer expenditureson communications and transportation than either of those permitted. The conceptual, andpartly empirical, decomposition of those trends explored here enhances our understandingof these important expenditure categories, and provides a useful methodology with which toexamine trends in other categories as well.
The organization of the rest of this paper is as follows. In Sect. 2, we analyze at a con-ceptual level the sources of change over time in an individual’s expenditures on a givencategory of goods. In Sect. 3, we describe the data available to our empirical analysis: theU.S. Consumer Expenditure Survey and the Consumer Price Index series. Section 4 presentsthe results, starting with trends in the raw data, and then successively accounting for thevarious sources of change. Section 5 concludes the paper.
2 A conceptual decomposition of expenditure trends
The expenditure on goods category i , Ei , can be expressed with the simple and well-knownequation Ei = pi qi , where pi is the unit price of i and qi is the quantity (number of units)demanded of i . Our straightforward question is, why does an individual’s expenditure oncategory i change over time? Here we focus on initial spending changes within a categorydue to an external stimulus related to that category, holding all else constant. However, ofcourse, an externally-stimulated change in (for example) the price of i (pi ) will have notonly the “instantaneous” effect on Ei , holding qi constant, but also follow-on effects on thedemand for i and other goods j (qi and q j ) via own- and cross-price elasticities (and sim-ilarly, changes in the prices of other categories j will affect the demand for i , i.e. producesubstitution and complementarity effects, through cross-price elasticities). These changes inq will in turn affect p, and so on. By the same token, an externally-stimulated change in thedemand for i (qi ) will have an instantaneous effect on Ei , but also ripple effects on ps andthence qs.
123
A decomposition of trends in U.S. consumer expenditures 3
quality change ( similar item)
production change (same item)
technologicalchange
income change
taste change
Fig. 1 Relationships among influences on consumer expenditures
A number of partly overlapping reasons for changes in Ei present themselves, which canbe labeled inflation, production changes, technology changes, quality changes, taste changes,and real income changes (Krüger 2008 presents a similar list of factors influencing aggre-gate structural change in the economy). We discuss each reason in turn, relating them to theequation for Ei . Relationships among these reasons are illustrated by the Venn diagram inFig. 1.
1. Identical goods are bought, but they cost more due to inflation (pi increases). Thus, ifa given basket of goods costs more due to inflation but income has not increased to “keepup with inflation”, one will spend more (not only in nominal dollars but as a proportionof income) to buy the same basket of goods in category i (and generally spend less inanother category).
2. Identical goods are bought, but they cost less, or more, due to real changes in thecosts of inputs, production, and/or distribution (pi changes). We will loosely refer tothis case as production changes. Examples include:
• increased production efficiencies result in lower costs;• cheaper labor is found, which lowers the cost;• material inputs (e.g. minerals, petroleum products) become more scarce and therefore
more costly, raising the price of the finished good;• transportation of the finished good becomes more (or less) expensive, raising (or
lowering) its retail price (e.g., Hummels 2007, p. 131 suggests that oceanic shippingcosts have experienced “adverse cost shocks” in the past 50 years, though these havebeen largely counteracted by technological improvements that decreased costs; airfreight costs, by contrast, have unequivocally decreased, by an order of magnitude).
3. The cost of identical or similar goods changes due to changes in technology thataffect either production costs (for identical goods) or quality (for similar goods, asdiscussed below) (pi changes). Thus, technological improvements constitute one typeof change in production. Production-related technological improvements can generally
123
4 P. L. Mokhtarian et al.
be expected to lower costs (see, e.g., Milgrom and Roberts 1990), and to the extent thosesavings are passed on the consumer, lower prices result. The same may not be true ofquality-related technological changes, as discussed below.
4. Previously-available goods are replaced or augmented by similar goods of differentquality (affecting pi at the “group” level, meaning for the group of similar goods, andq i at the “item” level). Obviously there is some ambiguity about the words “similar”and “quality”. Note that the change in quality may not be due to a change in technol-ogy—although it often is—and what constitutes “technology” can be equally ambig-uous. Goods produced by technological change often have better quality compared topreviously-available goods, but by no means always (e.g., where mass production fromcheaper materials renders them flimsier than their earlier counterparts). Even if the newgood is higher in quality than the old one, its price could be either higher (to reflect thehigher consumer value arising from the higher quality) or lower (if the higher qualityis achieved through technological advances that also lower costs) than that of the pre-vious one. If previously-purchased goods are completely replaced by similar ones ofdifferent quality, the consumer can no longer choose to purchase the previous goods, andwe would consider this change in expenditures to be purely due to the quality change.Conversely, if previously-purchased goods continue to coexist alongside the new goods,and the consumer chooses the new goods, then both changes in quality and changes intaste are at work. However, the boundary between these two cases is blurry, since it isarguably changes in taste that drive a different-quality item off the market after some(possibly brief) period of co-existence.
5. Consumers’ tastes or preferences have changed (affecting qi at the micro scale, andpi at the macro scale). At the micro level, taking availability as exogenously determined,individuals may simply prefer more or less of various items over time (e.g. they may stopsmoking, or take up skiing, or reduce the frequency with which they go to movies or eatout, or have children—which alters spending patterns in many ways). These changes inpreference may or may not be related to changes in quality, as mentioned above—noneof the examples just offered depend on changes in quality. Similarly, they may or maynot be influenced by changes in technology. At the macro level however, availability isendogenous. Some goods are phased out and others are introduced in response to aggre-gate changes in consumer tastes over time—often enabled or driven by quality changes,with or without technology changes.
6. Real income changes over time (affecting qi , or shares pi qi/E). If an individual’sincome increases in real terms, she will decrease expenditures on inferior goods, whileincreasing expenditures on normal goods. The use of expenditure shares controls forchanges in total expenditure overall (which constitutes “consumed income” and is gen-erally taken as a proxy for total income in consumer demand models), but shifts amongshares are still expected as a function of income (as can be seen from the standard con-sumer demand equation system, which models expenditure shares as a function of pricesand income, or total expenditures, from which the income or expenditure elasticity ofdemand is computed). However (to the extent that tastes are viewed as synonymous withrevealed preferences, as opposed to latent ones), those shifts can basically be seen as dueto changes in tastes, which are correlated with changes in income. In other words, anexpenditure shift is not driven purely by a change in income; rather, a rise (say) in incomeenables various taste changes to be realized (while a fall in income may force some lessvoluntary taste changes to be “revealed”, e.g. from steak to ground beef). Note that whileincome effects can therefore essentially be classified as a subset of taste changes, tastechanges can occur even when income does not change.
123
A decomposition of trends in U.S. consumer expenditures 5
In summary, expenditures could change over time if:
• the unit price for an identical good changes, due to
– inflation,– production changes, i.e. changes in the real costs of inputs, production, and/or distri-
bution, including those due to technological improvements, and/or– macro-scale changes in demand due to taste changes;
• the unit price for a “comparable” good changes, due to
– inflation,– quality changes (with or without technological changes), and/or– macro-scale changes in demand due to taste changes;
• the unit price changes from infinite to finite or from finite to infinite (representing theintroduction of a new good or phase-out of an obsolete one, respectively), due to
– quality changes, with or without– technological changes, and/or– macro-scale taste changes; or
• the quantity demanded at a given unit price changes, due to
– changes in tastes, with or without– changes in income.
It is worth noting in passing the obvious point that, given all these potential sources ofchange, which can act in different directions, it is quite possible that expenditures on a givencategory may appear to remain relatively stable (even in real terms, i.e. after adjusted forinflation), while in fact the quantity, quality, and/or content of goods purchased may havechanged. When the physical impacts of a good are important (as with the environmentalimpacts of transportation, for example), it is critical not to focus exclusively on monetarymeasures of their consumption.
Another straightforward but important observation is that the total impact on expenditureEi of a decrease in real price (pi ) and an increase in demand (qi ) is ambiguous: the net direc-tion depends on the direct price elasticity of the good in question, as well as on the strengthof the taste changes that may also be driving the shift in qi (in addition to the price change). Ifprice changes are the major effect, then if demand is inelastic (direct price elasticity less thanone in magnitude) the change in Ei will tend to be negative: the decrease in price is not com-pletely counteracted by the increase in quantity. Conversely, if demand is elastic, the changein Ei will tend to be positive. However, these effects will be modified by taste changes and,for expenditure share, also by changes in total real income (the denominator of the share).
3 Data description
3.1 The U.S. consumer expenditure data
The Bureau of Labor Statistics (BLS) of the United States Department of Labor (U.S. DOL)collects consumer expenditure data (see U.S. DOL 1997, 2004) on a quarterly basis for largerand/or less frequent outlays (e.g. automobiles, mortgage payments) from a sample compris-ing more than 30,000 interviews a year, and on a weekly basis from a sample of around7,700 consumer units (loosely equivalent to a household). Expenditures are classified into
123
6 P. L. Mokhtarian et al.
seven major categories of items, and from there into further subcategories (http://www.bls.gov/cex/csxgloss.htm, accessed August 24, 2007).
From 1984 onward, consumer expenditure data are available on the BLS website (http://www.bls.gov). Accordingly, we assembled the annual data for the 19 years from 1984 to2002 (the latest year available at the time). For our detailed analysis we selected the sub-categories most relevant to transportation and communications, but defined those two termsmore broadly than usual so as to obtain a complete picture of expenditures in those areas.As a result, however, some of our subcategories contain items in addition to those related totransportation and communications. In particular, the “out-of-town lodging” subcategory isidentified as a transportation item in our context, because lodging away from home is likelyto be associated with transportation and as such would be part of the total cost of the trip, butit also includes expenditures on college dormitories. The “other entertainment equipment”subcategory is included under transportation since it contains bicycles and a number of otherrecreational travel vehicles, but also other less relevant items. And the “miscellaneous house-hold equipment” group is identified as communications since that is where computers andtelephone equipment are classified, but includes numerous other items as well.
The 14 groups (nine for transportation and five for communications) used in this studyare defined in Table 1.
3.2 The U.S. consumer price index data
The consumer price index (CPI) is a measure of the average change in the prices of a “mar-ket basket” (i.e. a representative sample) of consumer goods and services. The CPI is mostoften used as a measure of inflation and a deflator of other economic series (to produceinflation-adjusted series). The data most appropriate for our purposes is the CPI-U seriesrepresenting urban consumers (consisting of about 87% of the U.S. population in 1990; dataare not available for the entire population of both urban and rural consumers), published bythe BLS every month and available at http://www.bls.gov.
Like the Consumer Expenditure data, the CPI data is available for a number of categoriesand subcategories of items. Of course, the hierarchies of the two sets of data do not match.Furthermore, although CPIs for most categories are available for the study period (1984–2002), some were not available for the entire period, or the category changed over the courseof the period. Accordingly, it was necessary to manipulate the CPI data somewhat to developa complete set of measures for each consumer expenditure group shown in Table 1. Thedetails of that process are laid out in Choo et al. (2006).
CPIs are relative measures, requiring a base year (1984 in our case), whose CPIs are fixedto 100. For example, when the CPI-U for transportation for 2002 is 147.4 with a base yearof 1984, this can be interpreted that a bundle of transportation expenses costing $100 in1984 would have cost $147.4 in 2002, indicating an inflation rate of 47.4% for the period.Conversely, dividing the 2002 raw (“current” or “nominal” dollars) expenditure on transpor-tation by the CPI of that year and multiplying by 100 expresses the expenditure in “constant”(1984) dollars, and thus theoretically controls for inflation. In reality, however, as explainedfurther in Sect. 4.2, the effects of inflation are confounded with those of technological andquality advancements (e.g. one gallon of gasoline can take an automobile farther today thanit could in 1984), changes in consumer tastes (despite the availability of more fuel-efficientcars, consumer preferences for larger truck and sport utility vehicles mean that a gallon ofgasoline actually produces fewer miles traveled today than in 1984), and occasional changesin the market basket to reflect new (e.g. computers) and obsolescent (e.g. record albums)goods and services.
A decomposition of trends in U.S. consumer expenditures 7
Table 1 Expenditure groups used in this study
Group Description
Transportation
Vehicle purchases:cars and trucks, new
The purchase of new domestic and imported cars and trucks and other vehicles,including motorcycles and private planes
Vehicle purchases:cars and trucks, used
The purchase of used domestic and imported cars and trucks and other vehicles,including motorcycles and private planes
Vehicle financecharges
The dollar amount of interest paid for a loan contracted for the purchase ofvehicles (new or used, domestic or imported, cars and trucks and other vehicles,including motorcycles and private planes)
Gasoline and motoroil
Gasoline, diesel fuel, and motor oil
Vehicle maintenanceand repairs
Tires, batteries, tubes, lubrication, filters, coolant, additives, brake andtransmission fluids, oil change, brake work including adjustment, front-endalignment, wheel balancing, steering repair, shock absorber replacement, clutchand transmission repair, electrical system repair, exhaust system repair, bodywork and painting, motor repair, repair to cooling system, drive train repair,drive shaft and rear-end repair, tire repair, audio equipment, other maintenanceand services, and auto repair policies
Vehicle insurance The premium paid for insuring cars, trucks, and other vehicles
Other entertainmentsupplies, equipment,and services
Indoor exercise equipment, athletic shoes, bicycles, trailers, purchase and rental ofmotorized campers and other recreational vehicles, camping equipment, huntingand fishing equipment, sports equipment (winter, water, and other), boats, boatmotors and boat trailers, rental of boats, landing and docking fees, rental andrepair of sports equipment, photographic equipment and supplies (film and filmprocessing), photographer fees, repair and rental of photo equipment, fireworks,and pinball and electronic video games
Out-of-town lodging All expenses for homes, school, college, hotels, motels, and other lodging whilepeople are out of town (primary residence expenses are included elsewhere andnot analyzed here)
Public transportation Fares for mass transit, buses, trains, airlines, taxis, school buses for which a fee ischarged, and boats
Communications
Telephone service All charges related to telephone calls
Miscellaneoushouseholdequipment
Typewriters, luggage, lamps and light fixtures, window coverings, clocks,lawnmowers and gardening equipment, other hand and power tools, telephoneanswering devices, telephone equipment and accessories, computers andcomputer hardware for home use, computer software and accessories for homeuse, calculators, business equipment for home use, floral arrangements andhouse plants, rental of furniture, closet and storage items, other householddecorative items, infants’ equipment, outdoor equipment, smoke alarms, otherhousehold appliances, and other small miscellaneous furnishings
Television, radios,sound equipment
Television sets, video recorders, video cassettes, video tapes, discs, disc players,video game hardware, video game cartridges, cable TV, radios, phonographs,tape recorders and players, sound components, records, compact discs, andtapes, musical instruments, and rental and repair of TV and sound equipment
Postage and stationery All kinds of postage and stationery supplies
Reading Subscriptions for newspapers and magazines; books through book clubs; and thepurchase of single-copy newspapers, magazines, newsletters, books, andencyclopedias and other reference books
Source: Based on http://www.bls.gov/cex/csxgloss.htm. Accessed 24 Aug 2007
We decompose the expenditure trends in several stages, with the results summarized in Fig. 2.In Sect. 4.1, we review the most basic trends: those for the raw expenditures in each group.In Sect. 4.2, we adjust the raw data for inflation, control for changes in real income by exam-ining (CPI-adjusted) expenditure shares, and present the effects of changes in income ongroup-specific expenditure shares, as well as the direct price elasticities for completeness.These are the influences we can empirically distinguish (to some extent). In Sect. 4.3, wespeculate with respect to the remaining effects: those of production, technology, quality, andtaste changes.
4.1 Trends in the raw data
Figure 3 shows the nine transportation and five communications expenditure trends. Expen-ditures in all 14 groups are generally increasing over time, although several groups (e.g.Veh_New (new vehicle purchases), Gasoline, and Oth_Lodg (out-of-town lodging)) are some-what erratic, and the Reading category is relatively flat (being the only one whose time trendis not significant at the 0.05 level). These results are reflected in the second column ofFig. 2.
Among the transportation categories, Veh_Used (used vehicle purchases) shows the largestrate of increase ($64/year, on average). The Veh_New (new vehicle purchases) and Veh_Insur(vehicle insurance) categories also show relatively high rates of increase ($27/year each),indicating that private vehicle related expenditures increased faster than those for public trans-portation ($11/year), out-of-town lodging ($5/year), or other entertainment (including recre-ational vehicles, $12/year). Among the communications categories, Telephone and Misc_HH(miscellaneous household equipment, including computer software and hardware) showrelatively high rates of increase ($29 and $27/year, respectively) and TV_Radio shows amoderate increase of $18/year, while Postage and Reading show increases of only $2/yearand $0.54/year, respectively. These results are consistent with expectations, in view of thedramatic advances in telephone and computer-related technology in recent years.
4.2 Trends adjusted for inflation and changes in total income; direct price elasticities
In Sect. 2, we discussed a number of potential reasons why consumer expenditures on acertain category change over time. In the process, it should have become clear that an exactdecomposition of changes in consumer expenditures is unfortunately impossible, due to ambi-guities in the sources of change, boundaries and overlaps between the various sources, andinadequate data on each source. It is possible, however, to control for the effects of inflationto some extent, using the CPI which has been devised for primarily that purpose.
In reality though, CPIs are not a pure measure of inflation. The main reason is that the“market basket” of goods used to establish the CPI changes over time. The goods generallychange somewhat each year (with about 30% of goods being replaced each year according toMoulton and Moses 1997), reflecting availability, quality, and/or minor technology changes.From time to time, altogether new goods are added (either supplemental to or in replacementof older goods), reflecting major technology changes (e.g. cellular phone service was addedin 1998). As argued by Deaton (1998), and demonstrated by Hausman (1999) and Bils andKlenow (2001), the BLS-calculated CPI overstates the rate of inflation by failing to com-pletely control for quality and technology changes in the market basket. In other words, part
123
A decomposition of trends in U.S. consumer expenditures 9
of an increase in prices attributed to inflation is actually due to quality improvements and theintroduction of new technologies (Durand 2007). As it is a matter of scholarly debate andbeyond the scope of this study to correct for that bias, we must accept that our CPI-adjustedexpenditures control not only for inflation, but to some unknown extent for quality changesand some technology and taste changes as well. Nevertheless, for the sake of discussion wewill suppose that the adjustment is mainly controlling for inflation.
Figure 4 shows the trends in the CPI series corresponding to our expenditure groups.CPIs that rise over time indicate categories that have been subject to inflation, while CPIsthat remain relatively stable indicate categories that have escaped inflation. We ascertainedthe CPI trend for each category through simple regressions of the CPI series against year(not shown): the sign of the coefficient of year in the regression indicates whether the CPIbasically rose, fell, or remained stable (on average) over time.
The third column of Fig. 2 summarizes those results, showing that unit prices were subjectto inflation in all but two of our 14 categories. Vehicle finance charges were fairly erratic butshowed no significant time trend, while the miscellaneous household goods group showed amarked decline in prices over time. The construction of the latter CPI series (actually based oninformation technology hardware and services from 1998 onward, and information process-ing equipment before 1998) and its implications are discussed at some length in Choo et al.(2006). Here, suffice it to say that, probably more than any other, this series is confoundingtechnology, quality, and taste changes with those due to inflation.
Figure 5 portrays the trends in expenditure shares, expressed in real (CPI-adjusted) dol-lars. The use of real dollars controls for inflation, and the use of shares controls for increasesin total expenditures (income). These trends, then, provide a more informative overview ofexpenditure patterns in each category, and will be the starting point of the remaining discus-sion. (The adjusted Misc_HH series is not reliable due to a mismatch of categories and otherreasons explained in Choo et al. 2006, and since it takes on a much wider range of valuesthan the other series do, plotting them all on the same figure makes it difficult to analyzetrends in the other 13 expenditure groups. Accordingly, we have dropped the Misc_HH groupfrom the figure here, although it is available in the accompanying report.) In dramatic con-trast to the trends in raw expenditures and in unit prices due to inflation, only six of these14 trends are positive. Thus, for the majority of categories (including new vehicles, all threevehicle operating cost groups—gasoline, maintenance, and insurance—out-of-town lodging,public transportation, and the two “old technology” communications groups of postage andreading), spending as a share of income has actually declined over time, in real terms.
As a matter of interest, Table 2 compares the trend directions for four different forms ofexpenditures (the regression equations on which the table is based are presented in Chooet al. 2006): the levels in current dollars (as shown in Fig. 3), shares of total expendituresin current dollars, levels in CPI-adjusted (constant) dollars, and shares of total CPI-adjustedexpenditures (Fig. 5). It can be seen from the table that the latter three sets are very similarto each other with respect to trend directions for each expenditure group (and quite differentfrom the raw trends of the first set), indicating that (since much of the increases in incomeare due to inflation) controlling for increases in income also tends to control for inflation.The fourth column of Fig. 2 recapitulates the trends for the fourth set—expenditure shares,expressed in real dollars.
However, there has also been non-negligible growth in real total expenditures over thestudy period, from $21,975 in 1984 to $26,102 in 2002 (1984 dollars). Since these figures areon a per-consumer-unit basis, and since household sizes have declined over the same period(from 2.71 in 1984 to 2.58 in 2002, U.S. Census Bureau 2005), it is clear that increasesin per capita consumption are even more pronounced. As discussed in Sect. 2, changes in
123
10 P. L. Mokhtarian et al.F
ig.2
Sum
mar
yof
estim
ated
and
post
ulat
edm
ajor
effe
cts
onex
pend
iture
sfo
rea
chgr
oup
Cat
egor
y R
aw E
x-pe
ndit
ure
Tre
nd1
Infl
atio
n(C
PI
Tim
e T
rend
)2
Exp
endi
ture
Shar
e T
rend
in
Rea
l $3
Inco
me
Cha
nge4
Pri
ceC
hang
e5
Tec
hno-
logi
cal
Cha
nge
Pro
duc-
tion
Cha
nge
Qua
lity
Cha
nge
Tas
te
Cha
nge
Dir
ectly
af-
fect
ed f
acto
r p i
q i
p i
p iq i
/ E
p i
q i /
E
p i
p i
p ip i
(gro
up)
q i(i
tem
) q i
(mic
ro)
Tra
nspo
rtat
ion
Oth
_Ent
er+
+
+
+ (L)
+ (L)
E
II
–
–
+
+
Oth
_Lod
g +
+
–
+ (L)
II/ E
–
– –
–
Pub_
Tr
+
+
– + (L
/N
)II
–
(air
) –
(air
)
+
(air
) –
(tra
nsit)
Veh
_New
+
+
– +
(L
) II
–
+
+
Veh
_Use
d +
+
+
–
+
+
Veh
_Fin
+
0 +
(p=
0.08
) –
–
+
G
asol
ine
+
+
– +
(N
) U
/I
–
+
Mai
nt+
+
–
–
+
Veh
_Ins
ur+
+
–
– –
+
Com
mun
icat
ions
T
elep
hone
+
+
+
+ (
N)
U
– –
– +
+
M
isc_
HH
+
–
+
–
– +
+
T
V_R
adio
+
+
+
–
– +
+
Po
stag
e+
+
–
+ (
~U)
I
+
– R
eadi
ng0
+
–
–
Not
es:
Em
pty
cells
may
refle
ctun
know
nor
ambi
guou
sef
fect
s,no
tne
cess
arily
noef
fect
.Po
sitiv
eor
nega
tive
sign
sin
the
q ico
lum
nsof
the
qual
itych
ange
and
tast
ech
ange
cate
gori
esdo
notm
ean
that
quan
titie
sof
alli
tem
sin
the
cate
gory
incr
ease
orde
crea
se,r
espe
ctiv
ely.
Rat
her,
apo
sitiv
esi
gnm
eans
that
the
nete
ffec
tof
incr
easi
ngth
equ
antit
yof
good
spu
rcha
sed
inth
atca
tego
ry,a
nd/o
rsh
iftin
gpu
rcha
ses
from
low
er-p
rice
dgo
ods
tohi
gher
-pri
ced
ones
inth
eca
tego
ry,i
sto
incr
ease
real
expe
nditu
resh
are
for
that
cate
gory
(and
conv
erse
lyfo
ra
nega
tive
sign
)a
Bas
edon
the
time
coef
ficie
ntin
are
gres
sion
ofcu
rren
t-do
llar
expe
nditu
reag
ains
tyea
r(s
eeC
hoo
etal
.200
6fo
rth
eeq
uatio
ns)
bB
ased
onth
etim
eco
effic
ient
ina
regr
essi
onof
CPI
iag
ains
tyea
r(s
eeda
tain
Fig.
4)c
Bas
edon
the
time
coef
ficie
ntin
are
gres
sion
ofC
PI-a
djus
ted
expe
nditu
resh
are
agai
nsty
ear
(see
Tabl
e2)
123
A decomposition of trends in U.S. consumer expenditures 11
income are expected to lead to redistributions of the levels and shares of expenditure on agiven category. Through estimation of the demand system of equations, those income effects(expenditure elasticities) can be ascertained. We did not estimate a 14-equation model due tosample size limitations, but did estimate models for two, four, five, and six equations basedon various aggregations of groups within the two main categories of transportation and com-munications. Accordingly, the results shown in the fifth column of Fig. 2 are presented by themore aggregate category groupings used in the models (specifically, they are based on theexpenditure elasticity results from three model systems using five categories, and one usingsix categories; one of the five-category models is presented in Choo et al. 2007b, and all areprovided in Choo et al. 2006).
Not surprisingly, all income effects are positive, indicating that these categories com-prise normal goods rather than inferior goods. Some categories (entertainment, out-of-townlodging, and vehicle capital costs) are luxuries (elasticities greater than one), others (publictransportation, vehicle operating costs, and new communications) are necessities (elasticitiesless than one), and the old communications category, roughly speaking, has unit elasticity(depending on the model system, elasticities vary between 0.2 and 1.6). In any case, the con-clusion is that increases in income will increase real spending in all 14 of these categories, sonone of the negative trends in expenditure share seen in the immediately preceding columnare due to income effects.
The sixth column of Fig. 2 summarizes the dominant own-price elasticity results fromthe same model systems as for the expenditure elasticities. Recall that price changes canoccur directly via changes in technology, production, or quality, and indirectly via taste-driven demand changes. Interestingly, expenditure changes with respect to prices tend to beunit elastic or inelastic, with the main exceptions being for the entertainment, out-of-townlodging, and public transportation (including airline) categories.
123
14 P. L. Mokhtarian et al.
Table 2 Nominal and real trends in expenditures and shares for the 14-group classification
Personal vehicle transportation—capital Veh_New − * − *
Veh_Used + * + *
Veh_Fin − * +
Personal vehicle transportation—operating Gasoline − * − *
Maint − * − *
Veh_Insur + * − *
New communication Telephone + * + *
Misc_HH + * + *
TV_Radio + * + *
Old communication Postage − * − *
Reading − * − *
Note: * significant at 95% confidence level
4.3 Effects of production, technology, quality, and taste changes
Assessing the effects of production, technological, quality, and taste changes is necessarilyspeculative, since as indicated earlier it is not possible to separately identify them empiri-cally. However, informed speculation is certainly possible, and productive. Below, we discussgeneral categories (following the groupings in the second column of Table 2) with respect tothe likely impacts of such changes on their expenditure patterns.
123
A decomposition of trends in U.S. consumer expenditures 15
Other entertainment
Real expenditure shares for other entertainment (Oth_Enter), which includes rental of motor-ized campers and other recreational vehicles and boats, show a positive trend, indicating thatpeople are engaging in more (or more expensive) recreational activities across time. Clearly,as incomes rise, tastes change toward a more leisure-oriented society (Aguiar and Hurst2007). Production, technology and quality changes also come into play, as modern recrea-tional vehicles and equipment differ considerably in material content and capabilities, andenjoy greater production efficiencies, compared to a few decades ago.
Non-private vehicle
In the category of non-private vehicle expenditures, the negative trend for Oth_Lodg indi-cates that people are spending a smaller share of their real incomes over time for out-of-townlodging. Since leisure travel in general is increasing over time (positive income and tastechange effects; Meurs and Kalfs 2000), two possible explanations of this declining trend forout-of-town lodging are that (1) people are substituting into less expensive forms of overnightstays (e.g. staying more often with friends/relatives or in a hostel, rather than in a hotel (tastechanges), and/or (2) new entrants and competition in the lodging industry have lowered thereal unit prices of lodging (quality and production changes). We expect the second expla-nation to dominate, although there are signs that real prices (average daily room rates) areincreasing of late (Lomanno 2005).
For public transportation (Pub_Tr), the trend is also negative. Interpretation of this cate-gory is complicated because it contains airline travel as well as public transit. Together withHarford (2006), we expect that public transit—at least in the U.S.—is an inferior good (withexpenditures on it decreasing as income increases), and with Dargay and Hanly (2001), thatairline travel is a luxury good (having a positive income elasticity, greater than one). Perhapsnot coincidentally, the only two models we developed that isolate this category exhibit bothof those relationships: an income elasticity of 1.301 (denoting a luxury good) results whenusing one computation of composite CPIs, and one of −0.224 (denoting an inferior good,though not statistically significant) appears when using a different computation (see Chooet al. 2006 for a discussion of the two approaches). Thus, in terms of taste changes we needto distinguish between the two modes. Another factor in the declining trend of expenditureshares, however, must be the ongoing changes in the airline industry, with restructuring andcompetition from new entrants lowering the real “unit” price of travel over time (a productionchange; see, e.g., Gaynor and Trapani 1994). Accordingly, even as the quantity of air traveldemanded rises with income, expenditure shares increase at a slower rate due to lower unitprices. With respect to public transit, fares rose in real terms from 1985 till 1996, then fellthereafter (Pucher 2002).
Private vehicle capital
From Fig. 2 we can observe that the expenditure trend for new vehicles (Veh_New) is neg-ative and for used vehicles (Veh_Used) is positive. This means that people are increasingpurchases of used vehicles more than new vehicles across time, and now (as shown in Fig. 5)the two categories have almost equal expenditures.
The trend for new vehicles is probably the net of technology changes (lowering the costof producing “the same” car over time), quality (more amenities, even in “the same” car),
123
16 P. L. Mokhtarian et al.
and taste changes (generally migrating toward the larger, higher-quality end of the vehiclespectrum). For used vehicles, the same changes may be at work, but in differing proportions.Perhaps the increase for used vehicles is a result of lower-income people moving up theeconomic ladder and purchasing a first car (used), and/or even middle-income householdspurchasing second and third vehicles (e.g. for their teenage children) on the used car marketas their incomes rise (and tastes change to make multiple cars seem more of a necessity).Another possibility is that some people want to update their cars frequently so as to enjoytechnological and style innovations, but want to save money by purchasing used rather thannew cars.
The final category in private vehicle capital is vehicle financing (Veh_Fin). This trend isweakly positive, indicating that people are devoting larger shares of their incomes to thiscategory. This is presumably due to a change in tastes toward a greater acceptability of buy-ing on credit (see Kish 2006); it is apparently not due to rising interest rates over the studyperiod, since this category showed no significant trend in its CPI. The rise in expenditureshare occurs despite falling “production” costs due to the increasing role of the internet inselling financing, processing payments, and so on (however, these effects may not have takenhold by the year 2002 conclusion of our study period).
Private vehicle operation
All three items in the private vehicle operation category show decreases in real expenditureshares over time. The negative trend for Gasoline is especially interesting. Passenger-vehicle-miles traveled have steadily increased over time (BTS 2006), and simultaneously, consumertastes have shifted away from the compact, fuel-efficient cars made popular by the energy cri-ses of the 1970s and 1980s, toward larger, less efficient trucks and sport utility vehicles (U.S.EPA 2006). Thus, the decline in the gasoline expenditure share through 2002 has occurreddespite countervailing income and taste change effects, and therefore represents a substantialtechnological improvement (since all personal vehicles, even the larger ones, are lighter andmore fuel-efficient now than were their counterparts in the 1980s). Higher U.S. gasolineprices of recent years have reversed this trend, however, with demand showing a markedprice inelasticity (Hughes et al. 2008; Small and van Dender 2007).
The negative trend for vehicle maintenance (Maint) implies that people are spending lessmoney for repairing or maintaining their vehicles. Since they are driving more, spending moremoney on used vehicles, and less or the same on new ones (and hence might be expected tohave higher repair bills), the implication is that technological improvements have made vehi-cles more reliable and less maintenance-intensive over the years (see, e.g., Desai and Mehta1997, cited in Hunsaker 2000). Similarly, since the expenditure shares for vehicle insurance(Veh_Insur) have declined even though vehicle-miles traveled have actually increased overtime, the implication is that auto travel has gotten safer (Tables J-1 through J-4 of BTS 2006),due to a variety of factors such as better driver education, more experienced older drivers,anti-drinking-and-driving campaigns, and technological improvements in vehicles and road-ways. The resulting effect on insurance rates can be interpreted as a unit price decrease. Aportion of such a decrease may also be due to the economies afforded by doing insurancebusiness over the internet, as discussed for vehicle financing as well.
New communications
All three components of the new communications category show increasing real expenditureshares. The real cost of providing a unit of Telephone service has been stable-to-declining
123
A decomposition of trends in U.S. consumer expenditures 17
over time, as a result of technological advances and industry restructuring (Hausman 1999;Wolak 1996). At the same time, the increases in service quality that have come about dueto technological improvements have contributed to changing tastes, reflecting increasingdemand for the more advanced services that have become available over time. Similar con-siderations apply to miscellaneous household equipment (Misc_HH) including computerhardware and software, and to the TV and radio (TV_Radio) category. In the latter case,technological advances have improved the resolution of TV (e.g. DLP: Digital Laser Projec-tion) and the sound quality of radio (e.g. satellite radio, high density radio). Cable televisionhas increased consumer choice. Rapidly changing technologies (stereos, magnetic tape play-ers, digital devices) have to some extent forced consumers to keep updating their home audioand video equipment as well as content platforms (from records to compact disks, videotapesto digital video disks).
Old communications
Finally, in the old communications category, the negative trends for Postage and Read-ing probably largely indicate taste changes away from postal service and printed materials(newspapers, magazines, books, letters), as new communication methods (e.g. e-mail, onlinebusiness, internet news sites) have replaced many old ones. For example, Nie and Erbring(2000) found higher internet usage associated with lower readership of newspapers. Someresearch has found, to the contrary, a complementarity effect between internet use and news-paper readership (Stoneman 2008), but the rise of news consumption online is widely citedas a major factor in the demise and downsizing of many physical newspapers. Another factorin this trend is the rise in the “production” cost of postal service during the study period (thereal price of a first-class U.S. postage stamp rose 20%, from $0.20 in 1984 to $0.24 (1984dollars) in 2002 ($0.37 in current dollars)).
These negative trends appear to belie historical claims of constancy of expenditure onmass media. However, Dupagne’s (1997) review of this literature criticized the theoreticaldeficiency of those studies, indicating that they had little connection to traditional economicor consumption theory.
5 Summary and conclusions
This study analyzes temporal trends in consumer expenditures on nine transportation-related and five communication-related groups of items, using annual data from the U.S.Consumer Expenditure Survey for the years 1984–2002. We conceptually decompose thetrends shown by the raw data into six partly overlapping sources: inflation, and changesin income, production, technology, quality, and tastes. We empirically identify the impactsof inflation by comparing trends in the raw expenditure levels for each group of items totrends in the corresponding CPI, and by expressing group-level expenditures as a shareof the total. We empirically identify the income effect using the expenditure elastici-ties from several different demand system models, representing different aggregations ofthe 14 groups of interest. And we qualitatively assess the impacts from the other foursources.
Very briefly, inflation nearly always increases unit prices (with exceptions for vehiclefinance charges and miscellaneous household equipment, including computer hardware andsoftware). Income effects are positive for all categories, meaning that these are all normal
123
18 P. L. Mokhtarian et al.
goods, not inferior ones. Broadly speaking, we speculate that taste changes have contributedto increasing expenditures in most categories, with the exception of out-of-town lodging, thepublic transit component of the public transportation category, and the old communicationmedia categories of postage and reading. We suggest that production (including industryrestructuring) and technological changes have led to decreased unit prices in most categories(postage being one exception). In the private vehicle operations categories, technologicalimprovements dominate, so that expenditure shares are decreasing (or were until recently, inthe case of gasoline) despite increasing demand. Conversely, in the new media categories,taste changes dominate, so that expenditure shares are increasing despite production andtechnological improvements which lower prices.
This decompositional analysis offers an approach to more formally structuring and ana-lyzing conventional wisdom with respect to group-specific expenditure trends. The presentapplication helps deepen our insight into trends for the important and related categories oftransportation and communications. However, the methodology is applicable to other cate-gories as well.
References
Aguiar, M., Hurst, E.: Measuring trends in leisure: the allocation of time over five decades. Q. J. Econ. 122(3),969–1006. http://www.frbsf.org/economics/conferences/0603/mtl.pdf (2007). Accessed 27 Aug 2007
Bils, M., Klenow, P.J.: Quantifying quality growth. Am. Econ. Rev. 91(4), 1006–1030 (2001)Bureau of Transportation Statistics (BTS): Transportation Statistics Annual Report: December
2006. U.S. Department of Transportation, Washington, DC. http://www.bts.gov/publications/transportation_statistics_annual_report/2006/ (2006). Accessed 27 Aug 2007
Choo, S., Lee, T., Mokhtarian, P.L.: Relationships between U.S. Consumer Expenditures on Communica-tions and Travel: 1984–2002. Research Report UCD-ITS-RR-06-13, Institute of Transportation Studies,University of California, Davis, July 2006. http://pubs.its.ucdavis.edu/publication_detail.php?id=1046
Choo, S., Lee, T., Mokhtarian, P.L.: Do transportation and communications tend to be substitutes, comple-ments, or neither? The U.S. consumer expenditures perspective, 1984–2002. Transp. Res. Rec. 2010, 121–132 (2007a)
Choo, S., Lee, T., Mokhtarian, P.L.: Relationships between U.S. consumer expenditures on communica-tions and transportation using Almost Ideal Demand System Modeling: 1984–2002. Transp. Plan. Tech-nol. 30(5), 431–453 (2007b)
Dargay, J., Hanly, M.: The determinants of the demand for international air travel to and from the UK. Paperpresented at the 34th Universities’ Transport Studies Group conference, Edinburgh, Jan 2002. http://www.cts.ucl.ac.uk/tsu/papers/UTSGAIR2002.pdf (2001). Accessed 27 Aug 2007
Deaton, A.: Getting prices right: what should be done. J. Econ. Perspect. 12(1), 37–46 (1998)Desai, P.S., Mehta, S.R.: Automobile Leasing: Affordability, Convenience and Reliability. Purdue University
Krannert School of Management (1997)Dupagne, M.: A theoretical and methodological critique of the principle of relative constancy. Commun.
Theory 7(1), 53–76 (1997)Durand, R.: Should we adjust input prices for quality changes?. J. Econ. Soc. Meas. 32, 1–14 (2007)Gaynor, M., Trapani, J.M. III.: Quantity, quality and the welfare effects of US airline deregulation. Appl.
Econ. 26, 543–550 (1994)Harford, J.D.: Congestion, pollution, and benefit-to-cost ratios of US public transit systems. Transp. Res.
D 11(1), 45–58 (2006)Hausman, J.: Cellular telephones, new products, and the CPI. J. Bus. Econ. Stat. 17(2), 188–194 (1999)Hughes, J.E., Knittel, C.R., Sperling, D.: Evidence of a shift in the short-run price elasticity of gasoline
demand. Energy J. 29(1), 113–134 (2008)Hummels, D.: Transportation costs and international trade in the second era of globalization. J. Econ.
Perspect. 21(3), 131–154 (2007)Hunsaker, J.: Maintenance contracts for leased goods: their role in creating brand loyalty. Manag. Decis.
A decomposition of trends in U.S. consumer expenditures 19
Kish, A.: Perspectives on recent trends in consumer debt. Payment Cards Center Discussion Paper06–05, Federal Reserve Bank of Philadelphia. http://www.philadelphiafed.org/pcc/papers/2006/D2006JuneConsumerDebtCover.pdf (2006). Accessed 27 Aug 2007
Krüger, J.J.: Productivity and structural change: a review of the literature. J. Econ. Surv. 22(2), 330–363 (2008)Lomanno, M.V.: Room-rate growth sustainable, according to long-term trends. Hotel Motel Manag.
220(8):26 (2005). Text without graphics available at http://findarticles.com/p/articles/mi_m3072/is_8_220/ai_n15697570. Accessed 28 Aug 2007
Meurs, H., Kalfs, N.: Leisure and vacation: a forgotten travel market? In the Report of the 111th Round Tableon Transport Economics, Transport and Leisure, pp. 123–146. European Conference of Ministers ofTransport (ECMT), Paris (2000)
Milgrom, P., Roberts, J.: The economics of modern manufacturing: technology, strategy, and organization.Am. Econ. Rev. 80(3), 511–528 (1990)
Mokhtarian, P.L.: A typology of relationships between telecommunications and transportation. Transp. Res.A 24((3), 231–242 (1990)
Mokhtarian, P.L.: Telecommunications and travel: the case for complementarity. J. Ind. Ecol. 6(2), 43–57,special issue on E-commerce, the Internet, and the Environment. http://www3.interscience.wiley.com/journal/120133161/issue (2003). Accessed 3 May 2009
Mokhtarian, P.L.: If telecommunication is such a good substitute for travel, why does congestion continue toget worse. J. Transp. Lett. 1(1), 1–17 (2009)
Moulton, B.R., Moses, K.E.: Addressing the quality change issue in the Consumer Price Index. Brook. Pap.Econ. Act. 1997(1), 305–350 (1997)
Nie, N.H., Erbring, L.: Internet and Society: A Preliminary Report. Institute for the Quantitative Study of Soci-ety, Stanford University. http://www.stanford.edu/group/siqss/itandsociety/v01i01/v01i01a18 (2000).Accessed 3 May 2009
Pucher, J.: Renaissance of public transport in the United States?. Transp. Q. 56(1), 33–49 (2002)Small, K., van Dender, K.: Fuel efficiency and motor vehicle travel: the declining rebound effect. Energy
J. 28(1), 25–51 (2007)Stoneman, P.: Exploring time use: a methodological response to ‘web-use and net-nerds’. Inf. Commun.
Soc. 11(5), 617–639 (2008)United States Census Bureau: Average Population Per Household and Family: 1940 to Present. http://www.
census.gov/population/socdemo/hh-fam/hh6.xls (2005). Accessed 13 July 2005U.S. Department of Labor (U.S. DOL): BLS Handbook of Methods. Chapter 16, Consumer Expenditures
and Income. http://www.bls.gov/opub/hom/pdf/homch16.pdf. Chapter 17, The Consumer Price Index.http://www.bls.gov/opub/hom/pdf/homch17.pdf (1997). Both files Accessed 8 Feb 2005
U.S. Department of Labor (U.S. DOL): Consumer Expenditures in 2002. Annual Report Provided by U.S.DOL, Bureau of Labor Statistics, Report No. 974, February 2004
U.S. Environmental Protection Agency (U.S. EPA): Light-Duty Automotive Technology and Fuel EconomyTrends: 1975 Through 2006. Report No. EPA420-R-06-011, July. http://www.epa.gov/otaq/fetrends.htm(2006). Accessed 28 Aug 2007
Wolak, F.A.: Can universal service survive in a competitive telecommunications environment? Evidence fromthe United States consumer expenditure survey. Inf. Econ. Policy 8, 163–203 (1996)
