This chapter emphasizes the importance of this unique federal survey by looking in more depth into the broad spectrum of its use. The chapter begins with a discussion of how the Consumer Expenditure Surveys (CE) are used in the construction of budget shares for the Consumer Price Index (CPI). Subsequent sections highlight many other uses of the CE, including a discussion of the role it plays in the administration of certain federal programs as well as in policy analysis and economic research.
In 2002, the National Research Council described the essential role of the CPI:
The Consumer Price Index (CPI) is one of the most widely used statistics in the United States. As a measure of inflation it is a key economic indicator. It serves as a guide for the Federal Reserve Board’s monetary policy and is an essential tool in calculating changes in the nation’s output and living standards. It is used to determine annual cost-of-living allowances for social security retirees and other recipients of federal payments, to index the federal income tax system for inflation, and as the yardstick for U.S. Treasury inflation-indexed bonds. (National Research Council, 2002)
The Bureau of Labor Statistics (BLS) calculates this index by “observing prices for a sample of goods and services that consumers purchase, and then creating aggregate estimates of price change using average expenditure
budget shares from data that CE provides” (Casey, 2010, p. 1). BLS publishes indexes on a monthly basis for different categories of products and services. Casey (2010) provided an in-depth description of the use of CE data by the CPI program. Most of the particulars included in this section are based on that paper.
The CPI program currently produces four indexes. The CPI-U is the most comprehensive index, measuring price changes for all urban consumers.1 A second index, the CPI-W, restricts that target population to the subset of urban consumer units in which the majority of income is earned in wage-earning or clerical occupations. A third index, the C-CPI-U, has the same population coverage as the CPI-U. Unlike the CPI-U, however, it uses an index formula that accounts for changes in consumer spending patterns in response to changes in relative prices at all levels of index construction. A fourth index, the CPI-E, is an experimental measure that reflects the spending patterns of urban consumer units in which the reference person is 62 years of age or older.
Types of Data Required by the CPI
Currently, the CE provides the CPI with expenditure data for urban consumer units, along with the demographic information necessary to implement the coverage definitions of the indexes described above.
For the CPI-U, the CE must (1) allow the identification of urban consumer units and (2) support the construction of subnational CPIs. Additional information is required on sources of income, the percent of income from different sources, and the age of the reference person in the consumer unit, in order to construct the CPI-W and the CPI-E, respectively. Finally, information on the housing tenure of the consumer is necessary for calculating expenditures on the components of shelter cost. Although no other demographic information is required for the current set of CPIs, Casey (2010) states that CPI researchers find additional demographic data useful for constructing other experimental indexes and pursuing other research.
1The CPI-U does not include the spending patterns of people living in rural nonmetropolitan areas, farm families, people in the Armed Forces, and those in institutions, such as prisons and mental hospitals.
For almost all expenditure categories, the CPI requires net out-of-pocket expenses exclusive of any finance charges. The main exception to this rule is the requirement that the CE collect the (implicit) rental value of owner-occupied houses to construct the budget share of the CPI component “Owner’s Equivalent Rent.” Expenditures on major home appliances and certain household maintenance expenses for homeowners are also imputed from the expenditures of renters on these items. This is another reason why housing tenure is a critical demographic variable in the CE. The CPI-U does not require expenditure data for investments, life insurance, interest payment, charitable contributions, or business expenses.
Although the CE currently collects a limited amount of information on where consumers purchase goods and services, the CPI does not currently use any of these outlet data. Rather, the CPI program uses a separate survey, the Telephone Point of Purchase Survey (TPOPS) (Bureau of Labor Statistics, 2011f), to gather this information. However, the CPI program would be interested in expanding the outlet data collection on the CE to provide alternatives to the TPOPS that would be more accurate and better integrated with the expenditure data.
The CPI uses information on income (total income; income from wage-earning and clerical worker occupations) to classify each unit in or out of the CPI-W population. Other than this, the CPI does not require information on income or any of its components, including child support and alimony payments.
Although the CE survey covers all consumer units in the country, the CPI uses only information on urban consumer units. Regarding geographic breakouts, the top priority of the CPI program is to measure the “All-items, U.S. City Average” index with precision. In order to do that, the current sampling methodology and index construction techniques require the CE to provide reliable, accurate expenditure estimates for 38 geographic areas. However, the publication of indexes for the 38 areas is of secondary concern.
CPIs are published on a monthly basis for the country’s three largest
metropolitan areas (New York, Chicago, and Los Angeles), bimonthly for another 11 metropolitan areas, and semiannually (using six-month averages) for 13 additional metropolitan areas. Separate regional indexes are also published for urban areas in three size classes—metropolitan areas with populations greater than 1.5 million, metropolitan areas with populations less than 1.5 million, and all nonmetropolitan urban areas (separate indexes for nonmetropolitan areas are not available for the northeast and west regions). Because of smaller sample sizes for both consumer expenditures and prices, the expenditure breakdowns for these geographically based indexes are less detailed.
Except for calculation of the C-CPI-U, the CPI program requires only annual expenditure estimates from the CE. Annual expenditure estimates needed to calculate the CPI-U, CPI-W, and CPI-E indexes are estimated by averaging annual expenditure budget shares over two consecutive years. The C-CPI-U, on the other hand, uses the expenditure budget shares from adjacent months to calculate price change between the two months, although information from the prior 12 months is used to reduce variance.
Expenditure Category Detail
The key requirement of the CE from the CPI program—the requirement that is potentially most demanding—is the need for expenditure detail. The CPI program uses CE data to calculate expenditure budget shares for 8,018 “elementary indexes.” The 8,018 budget shares are derived by multiplying the 211 item strata by the 38 subnational areas for which budget shares are required (31 areas for the 27 cities for which individual indexes are published, with 3 for the New York Combined Statistical Area [CSA], 2 each for the Chicago and Washington-Baltimore CSAs, plus 7 regions by city-size strata). The current CE sample is too small to support independent estimation of 8,018 budget shares, however, so the budget shares for subnational areas are derived by combining expenditure data specific to the subnational area with expenditure data for a broader geographical area that contains the subnational area using composite estimation. Composite estimation “weights” the subnational-area-specific and broader-area estimates according to their precision. The greater the variance in a subnational-area budget share (the lower the variance of the broader-area budget share), the lower the budget share assigned to the subnational-area budget share in the composite estimate. Moreover, the 27 metropolitan areas for which indexes are published are also part of the more aggregated region-by-size-class indexes.
Consequently, the real constraint for using CE data in the CPI program is the need for national item budget shares of an acceptable precision and enough precision at subnational levels to support an acceptable composite-estimation procedure. It was difficult for the panel to infer exactly what this requirement is.
That CPI requirements are not strictly imposed is reinforced by the fact that the necessary precision to select “entry-level items” is honored in the abeyance: “CE does not currently meet this requirement and CPI must aggregate expenditures to the ELI [entry-level item]-Region level in order to have a large enough sample for each probability” (Casey, 2010, p. 10).
The CE data provide an overall picture of consumer expenditures for the nation. In doing so, the CE provides detailed data on very specific expenses not available elsewhere. Federal agencies and some state agencies use a wide range of these specific estimates to administer important programs. Although far from a complete enumeration, this section describes some important uses of the CE in federal programs. Much of this information was presented at the June 2010 CE Data User Needs Forum (see http://www.bls.gov/cex/duf2010agendafinl.pdf) and expanded upon when panel members conducted follow-on discussions. A summary of some of those discussions is in Appendix C.
The CE Provides Important Information on the Cost of Health Care
The Centers for Medicare & Medicaid Services in the U.S. Department of Health and Human Services is responsible for producing the National Health Expenditure Accounts. Among many purposes, these accounts allow for the tracking and projecting of health care spending by businesses, households, and governments. These accounts contribute to the discussion of who ultimately pays for health care in the United States and the burden borne by different sectors of the economy to finance health care into the future. These are critical issues for the country today.
The Health Expenditure Accounts obtain data from a number of different sources, requiring consistent data over time. The CE is the source of private health insurance expenditures paid by households for individually purchased insurance. There is no other available source of consistent data. Additionally, the Health Expenditure Accounts use data from the CE to estimate out-of-pocket expenses for major health services. These include expenses for health services not covered (including deductibles and copayments) by insurance and public programs such as Medicare and Medicaid.
It also includes payments into Health Savings Accounts. The demographics and income data collected in the CE allow analysis of these data by age of head of household and household income. Staff at the Centers for Medicare & Medicaid Services use aggregate estimates published by BLS and also the CE microdata for additional analysis. They use income and asset data from the CE for special analyses (Cowan, 2010).
Taxpayers Have an Easy-to-Use Deduction Based on CE Data
The Internal Revenue Service (IRS) calculates “optional sales tax tables” using CE data. Taxpayers filing a Schedule A have the option to deduct state and local sales tax in lieu of state and local income tax. Many taxpayers, particularly those in states without a state income tax, choose this option. Those taxpayers may keep sales receipts throughout the year and calculate the sales tax they paid. Alternatively, they can use the IRS-supplied “optional sales tax tables” or online sales tax calculator to determine their deduction.
The IRS uses CE data to calculate household estimated sales tax for these tables (and for the online calculator). IRS supplies BLS with state and local taxability data. BLS combines this information with CE data to calculate household-level sales tax estimates. BLS provides these estimates, along with variables such as household income and family size, back to the IRS. The IRS models these data variables to produce the “optional sales tax tables” by household income and family size (Lee, 2010).
CE Data Help Support Child Welfare
The Center for Nutrition Policy and Promotion (CNPP) at the U.S. Department of Agriculture publishes Expenditures on Children by Families, an annual report that estimates what it costs to raise a child from birth through age 17, broken down by household income. This publication provides an extremely valuable source of information associated with child welfare. States use it in determining child support guidelines and foster care payments. The CE provides the major source of data for this publication, including child-specific expenditures such as clothing purchased for children. CNPP staff also use CE data on general household expenditures, allocating a proportion of these expenditures to children based on other sources of data (Lino, 2010).
CE Data Contribute to the Measurement of Poverty
In 1995, the National Research Council of the National Academy of Sciences issued a report titled Measuring Poverty: A New Approach
(National Research Council, 1995). The report criticized the methodology used to make the official poverty measurement and recommended improved methodology based on CE data. This new Supplemental Poverty Measure uses actual expenditure data for food, shelter, clothing, and utilities to derive poverty thresholds that are compared to measurements of disposable income from the Current Population Survey. This Supplemental Poverty Measure is currently being computed in addition to the historical measure (Short, 2010).
Good policy is created on a foundation of high-quality information about available options and sustained by analyses of whether the policy achieves its intended effect. CE data are used extensively to evaluate policy and conduct applied research on a wide range of issues important to American households. The CE’s value is that it provides the “complete picture,” tying household demographics to data on the complete range of consumers’ expenditures and incomes. It is used extensively by economic policy makers examining the impact of policy changes on economic groups, and by businesses and academic researchers studying consumers’ spending habits and trends. This section presents examples of the crucial analysis and research that depend on data from the CE. The intent of this section is to illustrate the breadth and depth of research and policy analysis made possible through CE data.2
Effect of Taxes and Tax Rebates Examined Using CE Data
Effects of Possible Cap and Trade Regulation
The Congressional Budget Office (CBO) is often called upon to project the possible effects of pending regulations. Harris and Perese (2010) did this for the highly publicized and politicized Global Warming Pollution Reduction Program proposed in H.R. 2454. Using the CE data, they predicted how the proposed regulation might affect the purchasing power of households at different income levels. (Their presentation at the BLS Data User Needs Forum was not part of an official CBO projection.) Grainger and Kolstad (2010) pursued a parallel but separate effort to the CBO staff members, also using the CE data. After concluding that these indirect taxes would inequitably affect households at lower income levels, they proposed
2Authors of a number of these studies use the terms expenditures, consumption, and spending somewhat interchangeably.
Impact of Direct Taxes on the Cost of Living
Gillingham and Greenlees (1987) defined a cost-of-living index including direct taxes. They used CE data to approximate the “tax and price index” (TPI) at the household level from 1967 to 1985. On average, the TPI increased much more rapidly than a CPI-type index, but the impact of taxes was highly progressive. They also used the TPI to evaluate alternative methods for indexing the federal tax system and to study an indexed system historically, comparing indexation with the CPI to actual tax policy, a tax system with constant parameters, and an “exact” indexing scheme (Gillingham and Greenlees, 1990). They concluded that (1) the sequence of tax reductions implemented between 1967 and 1985 fell short of mimicking indexation, (2) wealthier households would have benefited relatively more than lower-income households from indexation, and (3) CPI indexation would not have completely eliminated bracket creep.
Effect of Added Gasoline Taxes
West and Williams (2004) looked at potential increases in gasoline taxes and the likely distributional effect of those increases. They used the CE data to incorporate behavioral responses to estimate a demand system that included other goods and services as well as gasoline. They recommended implementing a larger gasoline tax and then using those available funds to reduce labor taxes. Archibald and Gillingham (1981) used CE data to analyze the distributional implications of either gasoline rationing or a tax on gasoline. The use of a model developed by Archibald and Gillingham (1980) implies that the incidence of a tax or the benefit of rationing with a “white market” in coupons would be very progressive.
Effect of Taxes on Charitable Giving
Reece and Zieschang (1985), building on Reece (1979), who also used CE data, used CE data to estimate models of the impact of tax deductibility on the level of charitable giving. They used econometric techniques that addressed the complexity introduced by a progressive step function of marginal tax rates to obtain consistent estimates. They then used the estimated parameters to shed light on the impacts of four alternative tax policies on the level of charitable giving. Their results did not support the proposition that the alternative policies they considered would lead to substantial increases in the level of charitable contributions at the cost of relatively small losses in tax revenue.
As part of an economic stimulus program, tax rebate checks were mailed to American households during the summer of 2001. Did households use these rebates in ways that would help stimulate the economy? Exploiting the panel data aspect of the CE, Johnson, Parker, and Souleles (2006) found that households spent roughly two-thirds of their rebate checks during the first six months after receipt. This study was possible because of the addition of questions to the CE to collect information about the amount of the stimulus checks and when they were received.
A similar economic stimulus program was initiated in May 2008. To analyze the 2008 stimulus, questions were again added to the CE about the rebate checks, including a question about what the households explicitly did with the checks. Paulin (2011) found that 49 percent of recipients used the money to pay off debt, while 30 percent reported that they spent the money. Younger recipients were more likely to spend the rebate than were older recipients.
CE Data Lead to a Better Understanding of the American Household
Gender Makes a Difference
Can the relative contributions to running a household by the members of that household be measured? Does gender make a difference in the value of the contribution? De Ruijter, Treas, and Cohen (2005) used data from the CE to “value” some routine domestic tasks, and categorized those tasks as typically “male” or “female.” For example, doing laundry might be a typically “female” task, while mowing the lawn a typically “male” task. They “valued” these tasks by equating their value with the amount households spent when they outsourced those specific domestic services. They also examined how those expenditures differ by living arrangement.
In an examination of the effect of gender on certain purchasing patterns, Kroshus (2008) assessed how much was spent by households on commercially prepared food (as a percent of total food expenditures) by gender and marital status. Not surprisingly, households headed by unmarried men spend a higher percent of their food expenditures on commercially prepared food.
Age Makes a Difference
Fisher et al. (2007) used 20 years of CE data to examine financial characteristics of older adults related to their home. As individuals grow older, their homes become increasingly mortgage-free. Even though this usually means that home equity also increases over this time period, few older
homeowners take out equity loans. CE data provide a unique way to look at generational differences. Paulin (2008) compared young never-married adults in 2004–2005 with similar individuals who responded on the CE two decades earlier (1984–1985). In real dollars, the 21st century young people spent a greater percent of total expenditures on shelter, utilities, and education. They spent less on food, transportation, and apparel than their 1980s counterparts. Health care was relatively unchanged.
Weagley and Huh (2004) used the CE data to look at the dynamics of retirement and near-retirement status on leisure expenditures. Not surprisingly, they found a positive correlation between leisure expenditures with retirement, income, and education.
Race and Ethnicity Make a Difference
The CE data are an ideal source for research on consumption spending as it differs by household racial and ethnic compositions. García-Jiménez and Mishra (2011) examined the demand for meat and meat products and found significant differences among households. Their results showed that white households purchase less meat (especially chicken and seafood products) than do Hispanic households. African American households purchase more pork and chicken than do Hispanics.
Marriage and Cohabitation Make a Difference
Households headed by single mothers, and how their income and consumption changed as a group between 1993 and 2003, were studied by Meyer and Sullivan (2008) using the CE data. The authors reported that income fell sharply (16%) in the first couple of years and then began to rise (17%) over the rest of the decade. For consumption, the authors found a modest (7% to 12%) rise throughout the decade.
Hawk (2011) used the CE data to understand differences in consumption spending between single people and married couples in their twenties. He found that the per capita income of singles was significantly lower than their married counterparts, that married couples were more likely to be homeowners, and that singles spent more per capita on housing, apparel, food, and education. Singles also spent less on health care.
DeLeire and Kalil (2005) examined expenditures on children by the living arrangements of their parents. Using data from the CE, they concluded that cohabiting-parent couples spend less money on education and more on alcohol and tobacco than do married-parent couples. Cohabiting-parent families had spending patterns different from those of divorced single-parent households and never-married single-parent households.
The Life-Cycle/Permanent Income Hypothesis (LCPIH) is the standard economic framework for understanding household spending and saving decisions over time. While the model provides a number of testable implications, the primary predictions involve how households will choose to consume in response to changes in income. When income changes are anticipated, the model predicts that consumption will not change contemporaneously, as households base consumption decisions in each period on expected lifetime wealth as opposed to current income. Unexpected income shocks, which alter lifetime resources, will result in consumption changes. Income changes can also be delineated between transitory and permanent income changes. Transitory income changes (e.g., a single-year windfall or loss) will only have a small impact on consumption as these changes have a small effect on lifetime resources. Permanent income changes, which alter income in all future years, will have a much larger effect on household consumption.
The CE has long been the unique data source that has enabled researchers to test predictions of the LCPIH using a broad set of consumption measures. Attanasio and Weber (1995) found that using microdata containing all expenditure measures for each household dramatically alters the empirical findings of previous tests of the LCPIH. First, the authors found that whereas many prior studies using aggregate expenditure (e.g., national time-series) data yield results inconsistent with the LCPIH, using microdata to create aggregates across households in a way that is consistent with the underlying economic theory results in estimates that are consistent with the model. Second, whereas past studies that had limited consumption measures (in many cases, just food consumption) rejected the LCPIH, testing the model using the full set of household consumption available in the CE could not reject the model.
Subsequent studies using the CE focused on clearly predictable changes in household income to avoid the many potential statistical pitfalls that may arise when predicting income changes for households using econometric methods. Some studies continue to find results that are consistent with the LCPIH. For example, Hsieh (2003) found that Alaskan residents, who receive large, annual oil dividend payments each fall, the amounts of which are pre-announced earlier in the year, do not exhibit a change in consumption upon receiving these payments. However, other studies find estimates that reject the LCPIH. Parker (1999) found that household consumption increases in response to intra-year paycheck increases due to households hitting the maximum annual Social Security tax limit, after which they no longer pay Social Security tax for that calendar year. Souleles (1999) found that household consumption increases in response to income tax refund
receipts, although the refund amounts are known to households before the checks arrive. Stephens (2008) found that household consumption increases once vehicle loans are paid off even though the date and amount of the final payment are known in advance by households. Finally, using the CE Diary data, Stephens (2003) found that households increase their daily nondurable consumption when their Social Security checks arrive, in contrast to the predictions of the LCPIH.
Since the LCPIH models the decisions of individual households, households are able to insure themselves against bad outcomes, such as a job loss or disability, only through their own savings. An alternative model that is an important economic benchmark for understanding the amount of risk that households face is the model of full insurance. In this model, individual households are fully insured against their own household-level changes in income, although aggregate-level income changes will influence household consumption (e.g., a village that pools all of its resources in each year and then redistributes them across all households in the village).
Mace (1991) tested the full-insurance hypothesis by exploiting the panel feature of the CE to regress changes in household consumption on changes in both aggregate consumption and household level “shocks” (e.g., changes in income and employment status). She found mixed evidence in support of this benchmark depending on the choice of empirical specification, although the preponderance of the evidence favors the model. However, Nelson (1994) found that alternative methods of measuring key variables, including using more expansive measures of consumption and employment changes, consistently reject the full insurance model. Attanasio and Davis (1996) focused on the large, observable wage changes that occurred between groups, as defined by education level and year of birth, during the 1980s to test the full insurance model. They concluded that the full insurance model overwhelming fails to explain the large between-group changes in consumption found in the CE over the same period.
CE Data Help Measure Well-Being Across Households
Income Inequality Across Households
Heathcote, Perri, and Violante (2010) combined data from the CE, the Panel Study of Income Dynamics, the Current Population Survey, and the Survey of Consumer Finances to conduct a systematic study of cross-sectional inequality in the United States. They found both a continuous and sizable increase in wage inequality over the study period.
Krueger and Perri (2006) investigated welfare consequences of this growing inequality. The CE helped reveal that poor households do not measurably change their consumption in response to lower wages, but instead
increase their working hours. The authors then assessed “household welfare” consequences using a number of techniques. They concluded that about 60 percent of U.S. households face welfare losses, with the size of those losses ranging from 1 to 6 percent of lifetime consumption for different groups.
Life-cycle models of variability in household savings and wealth accumulation (with comparable socioeconomic configurations) have ascribed cause to such factors as risk aversion, preferences for work or leisure in later life, and income replacement rates. Bernheim, Skinner, and Weinberg (2004) used data from the CE and the Panel Study of Income Dynamics to evaluate these conclusions. Instead, they found the “empirical evidence therefore casts doubt on theories that rely on differences in relative tastes for leisure, home production, or work-related expenses to explain the variation in wealth at retirement” (Bernheim, Skinner, and Weinberg, 2001, p. 854). The authors concluded that these factors appeared to be outside the context of the life-cycle model.
Understanding Poverty and How to Measure It
Fisher et al. (2009) examined the financial well-being of households of older Americans. They first distinguished between the notions of “income” poor and “consumption” poor. The authors emphasized that it is important to understand these two poverty definitions and the populations they imply in order to effectively measure the success of various poverty programs. Using 20 years of CE data, they reported that the measure of “poverty” is cut by one-fourth if that measurement uses both income and consumption. Older households that are white, homeowners, and married and have a high school diploma are more likely to be “poor” using only the income definition and not the combined definition. That is because they have sufficient assets to raise consumption above the poverty threshold.
Potential Nutritional Barriers in Poor Families
Do some households have to choose between paying heating bills and buying food? Bhattacharya et al. (2003) used the CE to track expenditures on both food and home fuels. They found that households (both rich and poor) had to increase expenditures on home heating during particularly cold periods. The difference: Poor families decreased their expenditures on food by about the same amount as they increased expenditure on home fuels, but richer families made no change in food expenditures during these same periods. The authors concluded that social programs need to understand this phenomenon and provide special assistance during cold-weather periods.
Stewart, Blisard, and Jolliffe (2003) concluded that low-income households spend less on fruits and vegetables than other households. They based this conclusion on a study of expenditures on fruits and vegetables and how these expenditures correlate with income. More surprisingly, these same households do not purchase more fruits and vegetables when they have a positive change in income.
Health Care Expenditures
What do households purchase when they do not purchase health insurance? Levy and DeLeire (2008) asked this question and used CE data to try to answer it. They found that households without health insurance spend more (compared to insured households) on things such as housing, food, alcohol, and tobacco. The authors raised the possibility that these households may be uninsured because they spend a greater percentage of total income on basic needs.
Does Medicare eligibility reduce out-of-pocket health care expenditures for those individuals? Have those expenses been changing over time? Duetsch (2008) looked at out-of-pocket health care expenditures of persons whose age was 55–64 (Medicare eligibility is 65) and those 65–74. Using CE data from 1985, 1995, and 2005, she found health care expenses (as a percent of total expenses) increased over those 20 years, but not consistently in real dollars. Between 1985 and 1995, the younger (ineligible) group’s health care expenditures decreased 27 percent, while the older (eligible) group’s expenditures decreased by 18 percent. Between 1995 and 2005, the younger group’s health care expenses rose by 22 percent and the expenses of the older group rose by 9 percent. In both decades, the older group spent more overall on health care than the younger group.
CE Data Are Used to Examine Credit and Debt in American Households
CE data are a tool for examining credit constraints and their effect. Ekici and Dunn (2010) examined credit card debt in its relationship to consumption. They used a monthly survey of credit card use to impute credit card debt into the CE data. They found a negative correlation between debt and change in consumption. Specifically the authors showed that a $1,000 increase in credit card debt leads to a 2 percent decrease in consumption growth. They also examined credit card debt by various household characteristics.
Grant (2007) investigated whether lower borrowing rates of some groups were related to credit constraints or lower demand. He estimated credit constraints and showed how these constraints differ by various household characteristics. His work found that households headed by
young, college-educated individuals were the most credit constrained. He also found that an observed lower level of borrowing among African American households appears to be affected by demand rather than credit constraints.
CE Provides a Tool to Examine Industry-Specific Markets and Special Topics
The comprehensive nature of the CE data allows for analysis that is targeted to expenditures within specific markets.
The probability of leasing a car is increased for households that are older, white or Hispanic, college educated, living in the Northeast and Midwest, living in a large Metropolitan Statistical Area, not having teenagers, and having a higher income. These results were found by Fan and Burton (2005) as they looked at the demographics that lead to a decision to “buy or lease” an automobile. However, the authors indicated that these effects are diminished when one controls for the vehicle characteristics.
Are communication expenses in some way a substitute for transportation expenses? Models developed by Choo, Lee, and Mokhtarian (2007) showed that these two areas of expenditure have both substitution and complementary effects.
Expenditures on Technology
Yin, DeVaney, and Stahura (2005) built a conceptual model using CE data to estimate the amount of money households are likely to spend on computer hardware and software. They then provided implications for consumers and policy makers. Hong (2007) used the CE data to examine the “substitution” relationship between expenditures on the Internet and other entertainment goods. He found Internet expenses have an effect on expenditures of recorded music.
Charitable and Political Giving by Households
There is a U-shape relationship between charitable giving and household income, with households at both the lower and higher income ranges giving a higher percentage of their income to charity than middle-income households. James and Sharpe (2007) found this result as they used CE data to examine the distribution of charitable giving by household income. The authors found that the charitable givers in the lower income ranges
are proportionately older, low-income but higher asset households. Dehejia, DeLeire, and Luttmer (2007) found that individuals who contribute to religious organizations are better able to insure consumption against income shocks. James (2009) examined the characteristics of households that make political contributions. He used a decade of CE data from 1995 to 2005. His analysis showed that political contributions were positively associated with income, wealth, education, and well-being. Political giving was negatively correlated with being a single female and being nonwhite.
For over a century, the collection of consumer expenditures on the CE and its predecessor surveys has played an irreplaceable role in understanding the market basket of goods and services that consumers purchase. While providing budget shares for the CPI remains a vital reason for the collection of consumer expenditures, a number of prominent uses of these data have emerged since the inception of these surveys. When contemplating revisions to the CE, it is important to remember that the CE has three critical but diverse uses, all of which have great importance for U.S. society: the CPI, the administration of a diverse array of government programs, and research that provides insight into policy decisions such as the effects of tax or other economic stimuli.