Explaining the 30-year shift in consumer expenditures from commodities to services, 1982–2012

The last 30 years have seen a shift in the allocation of U.S. consumer expenditures from commodities to services. This article uses Consumer Expenditure Survey (CE) data and Consumer Price Index (CPI) “relative importance” and index data (1) to show that the shift has been driven by changes not only in price but also in quantity¹ and (2) to identify the particular categories of services driving the overall shift to services consumption. Focusing on absolute changes in per-household expenditures during the period 1984–2011, the article finds a 9.1-percent increase in the quantity of services and no change in the quantity of commodities. This trend has been driven largely by a considerable increase in owner-occupied shelter. The article also finds that the quantity of health care services has decreased, although the share of personal consumption expenditures (PCE) accounted for by health care services, as measured from 1959 to 2009 by the U.S. Bureau of Economic Analysis (BEA), has increased. This difference illustrates that PCE data account for third-party expenditures, while CPI and CE data do not.² Within commodities, the quantity of durable goods has increased, while the quantity of nondurables has decreased.

Focusing on relative changes in total expenditures for the 30-year period from December 1982 to December 2012, the analysis yields the same general result. There has been an increase in the total quantity of services at the expense of commodities—a shift primarily due to an increase in the quantity of shelter, in particular owner-occupied shelter. There has also been a considerable increase in the quantity of “other services” (a residual category not including shelter, transportation services, medical care services, and energy services), when “other services” are compared with commodities. Moreover, the decrease in the quantity of medical care services does not mean that consumers have consumed less medical care services; rather, it means that third-party expenditures have increased while out-of-pocket expenditures have decreased.

In short, CPI and CE data show that the shift to a services-based economy entails more shelter, while PCE data indicate that the shift is also driven by increases in expenditures on health care services. Within commodities, there has been an increase in the quantity of durables and a decrease in the quantity of nondurables.

Background

In 1968, economist Victor Fuchs observed that more than half of the employed population in the United States was working in the services sector and thus was “not involved in the production of food, clothing, houses, automobiles, or other tangible goods.”³ The U.S. economy, he argued, had become a “service economy.”

Fuchs’s analysis identified a fundamental change on the production side of the U.S. economy, a change in which the services sector had captured an increasing share of overall employment in the United States. But a similar shift was also occurring on the consumption side of the U.S. economy, with consumers allocating an increasing share of total expenditures to services and a decreasing share to commodities. This change was already underway at the time of Fuchs’s writing, and since then, the U.S. economy has not looked back.⁴

Detailed CE data show that, from 1984 to 2011, per-household nominal average annual expenditures on commodities increased from $10,292 to $18,408, an increase of 78.9 percent, and per-household nominal average annual expenditures on services increased from $9,643 to $26,732, an increase of 177.2 percent. These changes are part of a longer term trend in relative expenditures that has seen the share of total expenditures allocated to commodities decrease and the share allocated to services increase. For example, in the period 1947–1949, commodities made up 72.9 percent of total consumer expenditures, compared with 27.1 percent for services, as measured by historical CPI relative importance tables, which use CE data to compute the percent allocation of consumer expenditures across multiple categories of commodities and services. By 1968, these shares had changed to 64.2 percent and 35.8 percent, respectively. In 2012, the shares had reversed: consumer expenditures on commodities had dropped to 39.7 percent of total expenditures, with services accounting for the remaining 60.3 percent.⁵ U.S. consumers now spend more money on services than on commodities. (See table 1.)

PCE data, an alternative source of information on consumption patterns produced by the BEA, show a similar trend. One BEA analysis reveals that the “share of current-dollar PCE accounted for by services increased from 45.7 percent in 1959 to 67.7 percent in 2009.”⁶

This fundamental change on the consumption side of the U.S. economy was driven, in part, by changes in relative price. As the prices of commodities rose at a lower rate than did the prices of services, the fraction of total consumer expenditures allocated to commodities decreased as well. Indeed, from 1968 to 2012, the CPI for All Urban Consumers (CPI-U) for commodities increased by 377.3 percent, which is equivalent to an average annual rate of change of 3.6 percent. Over the same period, the CPI-U for services increased by 780.0 percent, which is equivalent to an average annual rate of change of 5.1 percent. (See table 2.) PCE data show a similar trend. The BEA analysis referenced above reveals that, from 1959 to 2009, “services prices increased at an average rate of 4.4 percent, nearly 2 percentage points more than the 2.6 percent change in goods prices.”⁷

Even though consumers faced larger increases in the price of services, they nonetheless purchased a greater quantity of services. This article dissects the implicit price and quantity components of total dollar expenditures to show that the shift in expenditures was driven by changes not only in price but also in quantity. Moreover, commodities and services are disaggregated into major component categories (defined by CPI special aggregate indexes) to examine their expenditure, price, and quantity trends over the last quarter century, and to determine their relative influence on the overall shift from commodities to services. Because of the importance of the shelter index in the analysis, the article focuses on per-household expenditures during the period 1984–2011 and on total urban consumer expenditures from December 1982 to December 2012 and finds that the shift in quantity from commodities to services during both periods was driven largely by an increase in the quantity of shelter.⁸ In addition, there has been a relative shift in quantity from commodities to “other services” (e.g., cable television and Internet services) and to energy services (natural gas and electricity). Finally, the composition of commodities consumption has seen a decrease in the quantity of nondurable goods (e.g., food and apparel) and an increase in the quantity of durable goods (e.g., appliances and televisions).

The sections below examine in greater detail the trends in price, quantity, and expenditures for commodities and services. The next section summarizes the CE and CPI survey design methodology to illustrate how the CE and the CPI can serve as sources of information on changes over time in consumer preferences for commodities and services available in the marketplace. The section that follows provides an overview, based on CE data, of changes in average annual expenditures on commodities and services and deflates these changes with the use of appropriate CPI indexes to present an overview of absolute changes in quantity. The analysis then compares commodities and services, and their major component aggregates, against each other to determine whether one has become relatively more (less) expensive than the other and whether any observed difference in price has resulted in relatively less (more) quantity of a category being consumed over the examined period.

Consumer preferences in the CE and the CPI

The CPI is a measure of the average change over time in the prices paid by urban consumers for a “market basket” of goods and services, that is, a sample of goods and services that consumers purchase for day-to-day living. Produced monthly by the Bureau of Labor Statistics (BLS), the CPI weights each item in the market basket on the basis of the amount of spending reported by a sample of families and individuals. Widely used as a measure of inflation, the CPI is based on data collected in surveys that also provide useful information on changes in relative demand among U.S. consumers.

Price indexes are designed to measure the change in expenditure necessary to purchase a constant-quality market basket of goods and services.⁹ The CPI has two primary inputs: prices and expenditure weights. The prices are collected from two surveys: the Commodities and Services (C&S) Survey and the Housing Survey.¹⁰ Expenditure weights, the second input to price indexes, are based on CE data collected by the U.S. Census Bureau for BLS.

The CE identifies the dollar amount people spend on a broad range of goods and services. About 14,000 1-week diaries and 28,000 quarterly interviews are collected from the current CE sample each year. Until 1980, the survey was conducted about every 10 years. Since then, it has been conducted on an ongoing basis, with tables published annually.

The CE data are used as the source of expenditure weights in CPI indexes, as well as the source of “relative importance” tables historically produced by BLS on an annual basis. These tables contain the percent allocation of expenditures across all categories of commodities and services purchased by urban U.S. consumers, and thus provide comprehensive information on the overall composition of U.S. consumption over time. Until January 2002, CE data were used to update CPI weights approximately every 10 years. The CPI switched to biennial updating in January 2002.

The relative importance tables make it possible to identify changes over time in consumer expenditure allocations across the broad range of commodity and service categories for which CPI indexes are calculated. For example, urban consumers in the United States allocated approximately 6.2 percent of total expenditures to food away from home (restaurant meals, vending machines, mobile food trucks, etc.) in 2001, compared with 5.7 percent in 2012. By contrast, urban consumers allocated 5.8 percent of total expenditures on medical care in 2001, compared with 7.2 percent in 2012.

Food away from home and medical care are specific categories of goods and services for which indexes and expenditure allocations are computed. These indexes fall within the traditional aggregation structure of the CPI, in which indexes are computed for 8 major expenditure categories, 70 expenditure classes, and 211 item strata.¹¹ In addition to the traditional aggregation structure, the CPI produces special aggregate indexes in which lower level indexes are grouped into special categories, such as “all items less food and energy” or “all items less medical care.” Two special categories, which together encompass the full set of CPI indexes, are those for commodities and services.

There is not necessarily a one-to-one match between the specific goods and services tracked in the CE and the specific goods and services priced in the CPI. Expenditure weights are produced for all 211 basic item-area index cells (the lowest level of index construction), that is, for categories of goods and services, not for individual goods and services. Thus, inferences about shifts in consumer demand can be made only about categories of items, not about specific items. For example, inferences can be drawn that there has been a shift in quantity from commodities to services, or from nondurables to durables, but not that there has been a shift, say, from magazines to a specific package of cable television services.

In sum, data collected for use in the CPI can be used to measure shifting consumer preferences for various types of commodities and services currently available in the marketplace. In particular, the CE provides data on how dollar expenditures have shifted among various categories of commodities and services. BLS publishes annual relative importance tables that convert these dollar expenditures into a percent allocation of expenditures across the many categories of goods and services purchased by consumers.¹² CE and CPI expenditure data implicitly contain a price component and a quantity component. In the analysis below, these components are identified and analyzed to examine absolute and relative changes in consumption.

Toward a service economy: the rise of shelter

This section examines absolute changes in per-household expenditures on, and quantity of, commodities and services. It also considers changes in expenditure and quantity for the major component categories of commodities and services associated with CPI special aggregate indexes. These special aggregates are shelter, owners’ equivalent rent, transportation services, medical care services, “other services,” energy services, durables, and nondurables.

CE detailed expenditure data containing information on average annual expenditures per household at the most granular levels of disaggregation were obtained for the years 1984 and 2011. CE data are not categorized under commodities and services, or under CPI special aggregate groupings that can be assigned to either commodities or services, but instead are aligned with the traditional CPI aggregation structure and include categories such as food, housing, apparel, entertainment, transportation, medical care, personal care, and miscellaneous. These detailed groupings of data were thus manually assigned to special aggregate categories and summed to obtain expenditures for commodities and services and their major component aggregates. (See table A–1 in the appendix for a detailed listing of CE items mapped to each CPI special aggregate category.)

The resulting expenditures were analyzed to obtain percent changes in expenditures. CPI indexes were used to deflate percent changes in expenditures and then derive absolute percent changes in quantity.¹³ These derivations can be obtained from basic calculations involving price and quantity. The percent change in absolute quantity is given by

[P₁Q₁ / P₀Q₀] / (P₁/P₀) = Q₁/Q₀,

where P₀ and Q₀ are price and quantity in 1984, and P₁ and Q₁ are price and quantity in 2011. The product P₀Q₀ represents expenditures in 1984, and P₁Q₁ represents expenditures in 2011. The calculation of absolute quantity is the equivalent of deflating changes in expenditures by the change in the price index for the particular special aggregate category of the CPI examined. For example, from 1984 to 2011, average annual expenditures on commodities increased by 78.9 percent. Dividing this change by the change in the index for commodities yields a 0.4-percent increase in per-household quantity of commodities purchased, meaning that the per-household quantity of commodities was essentially unchanged. A similar calculation for services, for which average annual expenditures increased by 177.2 percent over the study period, yields a 9.1-percent increase in quantity after accounting for price change. Table 3 summarizes the results of these calculations for commodities and services, as well as special aggregate groupings of commodities and services.

Table 3 contains the changes in expenditure, price, and quantity for aggregate categories, such as transportation services and medical care services.¹⁴ Transportation services include such items as auto rentals, airline fares, and towing charges. Health care services include such items as physician services and hospital services. This study examines changes in out-of-pocket expenditures and quantity for aggregate categories, not for specific items within these aggregate categories. This distinction is important because the composition of the market basket (to which expenditures pertain) changed over the 1984–2011 period. Physician services in 1984 were different from physician services in 2011; so, an increase in “quantity” does not necessarily mean that consumers buy more of the same thing. Instead, the change in quantity is understood as a residual between changes in expenditures and changes in price and can be interpreted to mean that consumers buy “more” or “less” of a category of commodity or service, depending on whether the change in quantity is positive or negative. Consumers buy more or less of the same type of thing.

The components of services that show the most substantial increases in quantity are shelter and owners’ equivalent rent (OER). The quantity of shelter and OER increased by 18.5 percent and 20.2 percent, respectively, while everything else in services (services less shelter, services less OER) remained essentially unchanged.¹⁵ As explained in the next section, these increases are corroborated by increasing rates of homeownership and an increase in the median square footage of floor size of new single-family homes in the Unites States—an increase from 1,520 square feet in 1982 to 2,169 square feet in 2010.¹⁶

The quantity of “other services,” which contain CE items mapped to the CPI special aggregate category “other services” (e.g., telephone and Internet services; tuition and child care; admission fees to sporting and entertainment events; cable television; personal care services; and miscellaneous expenses, such as funeral expenses, legal fees, and checking account fees), decreased by 9.1 percent.¹⁷ Transportation services decreased by 5.5 percent. Energy services increased by 7.9 percent (as the purchase of shelter has increased, so has the purchase of natural gas and electricity).

Although the quantity of health care services decreased by 17.8 percent—a finding that may come as a surprise given an aging population and the growth of the health care industry—it is important to distinguish between total consumption expenditures and consumption attributed to consumers’ out-of-pocket expenditures. It is indeed the case that total consumption expenditures on health care services have increased over the long term. As explained in the BEA study on long-term consumption referenced earlier, “among PCE services categories, the largest increases in shares were for health care services and for financial services and insurance. The share of PCE accounted for by health care services increased from 4.7 percent in 1959 to 16.2 percent in 2009, while the share accounted for by financial services and insurance increased from 3.9 percent in 1959 to 8.1 percent in 2009. . . . Together, these categories accounted for more than two-thirds of the increase in the services share of PCE.”¹⁸ This increase in consumption expenditures for health care services was driven by large increases in employer-provided health care, government expenditures on Medicare and Medicaid, an aging population, and substantial advances in medical care technology.¹⁹

The BEA analysis on PCE trends also explains that, “in 1959, 72.7 percent of PCE for health care was paid for by direct out-of-pocket expenditures, and by 2009, this share had declined to 17.6 percent. During this period, the share of PCE for health care paid for by government increased from 3.4 percent to 45.4 percent.”²⁰ The health care marketplace in later years undoubtedly consists of much higher quality services that have benefitted consumers,²¹ but consumers pay much less out of pocket for such services, a trend highlighting a “significant decrease in the share of PCE accounted for by out-of-pocket expenditures.”²²

The CPI for financial services includes only checking accounts and other bank services, as well as tax return preparation and other accounting fees, and has less than 0.5-percent share in the CPI.²³ The PCE, however, is a more expansive category, which includes such items as pension funds, regulated investment companies (such as mutual funds), and securities commissions. According to the BEA article, “the increased share of financial services was associated with greatly increased holdings of regulated investment company—also known as mutual fund—assets by households, greatly increased use of credit cards, and large increases in the fees of banks and other depository institutions.”²⁴

Within commodities, the quantity of durables (e.g., appliances, personal computers, new vehicles, sporting goods)²⁵ increased by 39.0 percent, while the quantity of nondurables (e.g., food, clothing, and medical care commodities)²⁶ decreased by 9.9 percent. This indicates that the shift from commodities to shelter has been accompanied by a decrease in the quantity of traditional nondurable necessities, such as food and clothing, as well as an increase in the quantity of appliances, personal computers, new vehicles, sporting goods, and other durable goods. The PCE data also show a shift toward durable goods. From 1959 to 2009, the quantity of durable goods consumed increased at an average annual rate of 5.2 percent, compared with 2.5 percent for nondurable goods. Moreover, “the share (of PCE) accounted for by food and beverages decreased by 11.6 percentage points, from 19.4 percent to 7.8 percent, and the share accounted for by clothing and footwear decreased by 4.8 percentage points, from 8.0 percent to 3.2 percent.”²⁷

Table 3 highlights that the increased quantity of durable goods is associated with a small increase in the price of durable goods and that this small increase contrasts with the considerable increase in the price of nondurable goods. Expenditures on nondurables are still almost three times as large as expenditures on durable goods. In the next section, CPI relative importance figures show that durable commodities comprised only 8.8 percent of total consumer expenditures in 2012, compared with 13.1 percent in 1982; nondurable commodities comprised 30.9 percent of expenditures in 2012, compared with 39.9 percent in 1982. Consumers still spend more on nondurable items, such as food and clothing, but nondurables have become relatively more expensive and the quantity of durables has increased in the overall market basket of goods and services.

It is easily observed, then, that a considerable increase in the quantity of shelter has driven the overall shift from commodities to services. This finding contrasts with the findings of the BEA study of PCE data, which suggest that health care and financial services have driven the shift from goods to services. This difference in findings reflects differences in the scope and weight of measurement in the CPI and the PCE. The PCE weight on health care reflects third-party expenditures by employers and government, while the CPI does not.²⁸ The result is a higher weight on health care in the PCE than in the CPI, and a higher relative weight on shelter in the CPI than in the PCE. This difference in weight reflects the role of out-of-pocket expenditures in the CPI. Thus, the CPI measures consumption expenditures for which the cost is borne by consumers, whereas the PCE measures all consumption expenditures for the benefit of consumers.²⁹

The upshot is that the shift in quantity from commodities to services in the CPI has been driven largely by an increase in the quantity of shelter. More households have chosen to own homes than to rent. Those which own homes have tended to buy bigger homes, driving up demand for natural gas and electricity, trash and garbage collection, home appliances, and other items. Meanwhile, as total consumption expenditures have increased, out-of-pocket expenditures on health care services have decreased. In short, CPI and CE data show that the average household in the United States consumed relatively more shelter, as well as more durable items, in 2011 than in 1984, while PCE data show that the consumption of health care and financial services has increased even as out-of-pocket expenditures have decreased.

Toward a service economy: changes in the composition of the CPI market basket

In the previous section, CE data were used to obtain absolute changes in per-household expenditure and quantity. Expenditures were decomposed into price and quantity components to determine the overall increase or decrease in quantity of a specific category of consumer expenditure. In this section, CPI relative importance tables, which are constructed from CE data, are used to obtain relative changes in expenditures and quantity. The purpose is to examine changes in the composition of the CPI market basket by comparing commodities and services, and their major component aggregates, against each other, to determine relative changes in price and quantity.

Methodology. CPI relative importance data can be used to obtain the relative expenditure for a CPI category, such as commodities, at a specific point in time. Relative importance data for particular categories of expenditures are published as percentages of total expenditures, so that “all items” has a weight of 100 (i.e., 100 percent of total expenditures are spent on all items). These percentages can be interpreted as relative dollar expenditures, where a weight of 100 percent can be interpreted as $100 spent on all items. Thus, the relative importance of a CPI category in a “base” year can be interpreted as a scaled, or normalized, expenditure equal to P₀Q₀. The relative importance, or expenditure, in a later year is P₁Q₁.

These relative importances, or expenditures, can be divided into a price component and a quantity component. Holding quantity constant, the change in total expenditure resulting only from a change in price can be obtained. Similarly, holding price constant, the expenditure resulting only from a change in quantity can be obtained. Using these calculations, the implicit price and quantity components of a change in relative expenditure (i.e., relative importance) can be obtained, thereby isolating the real change in consumption (i.e., change in the quantity composition of the market basket) over time.

To obtain the expenditure based purely on price change, P₀Q₀ can be multiplied by the change in the index P₁/P₀ to obtain P₁Q₀. P₁Q₀ can be calculated for any two or more expenditure categories, such as commodities and services. The relative change in quantity for each category can then be obtained by dividing the expenditure in the later year (P₁Q₁) by the expenditure based on the pure price effect (P₁Q₀), where expenditures based on pure price change (P₁Q₀) are normalized so that shares sum to 100, and thus expenditures continue to be interpreted as shares of $100 spent on everything. Table 4 provides an example of these calculations for the relative shift from commodities to services.

For the CPI-U (urban) population, the share of total consumer expenditures attributed to commodities decreased from 52.9 percent in December 1982 to 40.0 percent in December 2012, while the share attributed to services increased from 47.1 percent to 60.3 percent over the same period. Based on the calculations described in the previous paragraph, it can be shown that if the quantity of commodities consumed was unchanged, so that any change in expenditures was caused only by a change in price, the share of consumer expenditures spent on commodities would have decreased to 42.8 percent, while the share spent on services would have increased to 57.2 percent (after normalization). This calculation indicates that the relative price (as opposed to the absolute change in price) of commodities decreased, while the relative price of services increased. In other words, in December 2012, consumers were able to buy more units of commodities per unit of services given up than they were in December 1982. Alternatively, in December 2012, consumers were able to buy fewer units of services per unit of commodities given up.

Based on this methodology, the change in relative quantity can be calculated as the percent difference between the normalized expenditure weights that would have prevailed under a pure price change and the actual expenditure weights from the CPI relative importance tables. Despite the decline in the relative price of commodities, the relative quantity of commodities turns out to have decreased by 7.2 percent, while the relative quantity of services has increased by 5.4 percent. This relative change in quantity was associated with a relative price decrease of 19.2 percent for commodities and a relative price increase of 21.5 percent for services. Although services have become relatively more expensive, the market basket contains relatively more services and relatively fewer commodities today than it did in December 1982.

The relative shift from commodities to services within the overall consumption basket purchased by consumers is unmistakable. But the commodities and services indexes are comprised of a large and diverse set of indexes, or commodity and service categories, such as food, apparel, energy commodities, rent of primary residence, owners’ equivalent rent of residences, transportation services, medical care services, energy services, and “other services.” Each of these indexes contributes to the all-items CPI, and, as such, attracts a certain percentage of the overall allocation of total expenditures by urban consumers. This percentage is the relative importance of the index. The following discussion explores which categories of services have driven the overall reallocation of expenditures and quantity from commodities to services.

Overview of changes in the relative importance of commodity and service categories. Figures 1 and 2 present the relative importance of major categories of commodities and services. The figures generally show that the expenditure shares of commodity indexes are decreasing, while the expenditure shares of services indexes are increasing. In figure 1, it can be observed that the relative importances of both durable and nondurable commodities decrease over the study period. The relative importances of consumer necessities such as food and apparel also decrease (the trend is harder to see for apparel, but apparel’s relative importance declined from 4.4 percent to 3.6 percent over the period).³⁰ “Core” commodities (i.e., “commodities less energy commodities”) and energy commodities both decreased in relative importance, although energy commodities were more volatile. The importance of groceries (food at home) declined from 12.9 percent to 8.6 percent, while that of food away from home declined from 6.1 percent to 5.7 percent.

In figure 2, the major standout is shelter, or more specifically, OER. Over the period, the services index and the “core” services index (i.e., “services less energy services”) both increase in line with the increase in OER (although there is a downward shift in shelter and OER beginning around 2008), with a larger blip observed from 1985 to 1986. The other larger blips occur in December 1997, when the relative importance of “other services” increased from 7.6 percent to 10.6 percent, while the relative importance of medical care services decreased from 6.1 percent to 4.4 percent. These changes were associated with the introduction in January 1998 of new expenditure weights based on the 1993–1995 CE. Thus, two indexes which appear to drive the increase in the relative importance of the services index are the indexes for OER and “other services.”

The impact of shelter. The index for shelter, which includes the indexes for “rent of primary residence” and OER, currently represents approximately 32 percent of all expenditures by urban consumers. Figure 2 reveals that, since OER’s introduction in the CPI-U, the relative importance of OER has increased from 13.5 percent of total expenditures in December 1982 to 24.0 percent in December 2012; over the same period, the relative importance of rent of primary residence in the CPI-U has increased from 6.0 percent to 6.5 percent (rent of primary residence increased from 5.9 percent to 6.5 percent in the last 2 years of the period, most likely as a result of the recent economic downturn).³¹ This shift means that the increase in the relative importance of shelter from 21.3 percent in December 1982 to 31.7 percent in December 2012 has been driven almost entirely by the change in the relative importance of OER. This interpretation is confirmed by figure 3, which shows the relative importances of OER and rent as shares of the overall relative importance of the services index.

View Chart Data

Shelter, OER, and rent as shares of the CPI-U services index, December 1982–December 2012 (in percent of services)

Year	Shelter	OER	Rent
1982	45.3	28.6	12.8
1983	45.3	28.6	12.8
1984	45.2	28.5	12.9
1985	45.6	28.7	13.0
1986	50.8	35.0	11.1
1987	51.1	35.4	11.1
1988	50.9	35.3	10.9
1989	50.8	35.4	10.8
1990	50.5	35.0	10.7
1991	50.2	34.8	10.5
1992	49.9	34.5	10.4
1993	49.5	34.3	10.2
1994	49.5	34.4	10.2
1995	49.6	34.5	10.1
1996	49.4	34.3	10.0
1997	51.9	35.2	12.0
1998	52.3	35.5	12.1
1999	52.3	35.4	12.2
2000	52.0	35.2	12.2
2001	53.7	37.6	10.9
2002	53.6	37.6	10.9
2003	54.9	39.0	10.3
2004	54.7	38.8	10.3
2005	54.5	39.6	9.8
2006	54.9	39.9	9.9
2007	55.5	40.8	9.8
2008	54.9	40.4	9.9
2009	53.7	41.9	9.9
2010	53.3	41.5	9.9
2011	52.5	39.9	10.8
2012	52.5	39.9	10.9
Source: U.S. Bureau of Labor Statistics.

The increase in OER’s relative importance from 14.0 percent in December 1985 to 19.1 percent in December 1986 (in figure 3, from 28.7 percent to 35.0 percent of services) deserves an explanation. The OER index was introduced in the CPI-U by BLS in January 1983,³² and its relative importance set to 13.5 percent. The relative importance was derived from actual CE data collected during the 1972–1973 CE, based on the following question: “If someone were to rent your home today, how much do you think it would rent for monthly, unfurnished and without utilities?”³³ The OER relative importance from 1972–1973 was then updated to December 1982 by relative price change, a procedure not unique to OER (the relative importances of other CPI categories for December 1982 would also have been based on the 1972–1973 survey data and updated by relative price change).³⁴ As stated by two BLS economists in June 1982, the weight for rental equivalence in December 1982 was “recalculated using the complex statistical estimating procedure used for weights in the official CPI.”³⁵

However, it must be noted that the relative importance of OER was 15.1 percent based on 1972–1973 CE data and 18.2 percent based on 1982–1984 CE data.³⁶ Thus, relative expenditures on OER increased, although the relative importance of 13.5 percent, which was obtained when OER was first introduced in December 1982, indicates that relative expenditures on OER declined. This result occurs because the 13.5-percent estimate reflects annual updating of the 1972–1973 relative importances by price indexes during the interval between the 1972–1973 and 1982–1984 updates of expenditure weights. Rents increased at a lower rate than the rate of overall inflation, so the relative weight of OER, as measured by price change, declined, even though the overall quantity of shelter services increased. In other words, the overall jump in the OER weight in December 1986 reflects a considerable increase in the relative quantity of OER, an increase that began in the 1970s. This is an example of how the CPI traditionally picked up major changes in quantity with a lag, namely, as part of major revisions, when new weights were introduced. These lags, and resulting data blips, have been reduced since 2002 with the introduction of new weights every 2 years instead of approximately every 10 years.

While the starting point for this analysis is December 1982, the trends in relative price and quantity should be understood as having been underway as early as the early 1970s. The December 1982 weight update reflects not only the date of OER introduction but also the factors driving the shifts in the 1970s in relative quantity from commodities to services. Indeed, demographic and economic changes in the 1970s help explain the increase in OER quantity and relative importance.³⁷

View Chart Data

Relative change in price and quantity for "rent of shelter" and "services less rent of shelter," December 1982–December 2012 (in percent)

Category	Price	Quantity
Commodities	-19.2	-7.2
Services	21.5	5.4
Commodities	-16.5	4.2
Services less rent of shelter	32.7	-5.2
Commodities	-10.8	-13.3
Rent of shelter	28.0	24.1
Source: U.S. Bureau of Labor Statistics.

The near doubling in the relative importance of the index for OER, and the accompanying large increase for shelter, warrants an analysis of how the relative shift in expenditures from commodities to services illuminates a relative shift in expenditures from commodities to shelter. Indeed, after accounting for price change, the increase in the quantity of shelter since December 1982 emerges as a major determinant of the shift in quantity from commodities to services. Comparing the indexes for commodities and “services less rent of shelter” reveals that the quantity of commodities increased by 4.2 percent, while the quantity of “services less rent of shelter” decreased by 5.2 percent. Comparing the indexes for commodities and “rent of shelter,” however, reverses the results: the quantity of commodities decreased by 13.3 percent, while the quantity of shelter increased by 24.1 percent. These changes are accompanied by a quantity decrease of 7.2 percent for commodities, compared with an increase of 5.4 percent for services. (See figure 4.)³⁸

These calculations are straightforward with the use of the officially published index for “rent of shelter.” The index is used to calculate the change in expenditure on shelter services that would result exclusively from a change in the price of shelter services. The change in quantity can then be derived from calculations described earlier.³⁹ Similar calculations comparing commodities with OER show that the quantity of commodities consumed decreased by 15.2 percent, while the quantity of OER consumed increased by 41.8 percent. Meanwhile, the quantity of commodities increased by 4.0 percent from December 1982 to December 2012, while the quantity of “services less OER” (an experimentally constructed index) decreased by 4.0 percent over the same period.⁴⁰ (See figure 5.)

View Chart Data

Relative change in price and quantity for OER and "services less OER," December 1982–December 2012 (in percent)

Category	Price	Quantity
Commodities	-19.2	-7.2
Services	21.5	5.4
Commodities	-17.9	4.0
Services less OER	28.1	-4.0
Commodities	-7.9	-15.2
OER	30.9	41.8
Source: U.S. Bureau of Labor Statistics.

The increase in the quantity of shelter purchased is corroborated by broad trends in the U.S. housing market over the last 30 years; these trends include (1) an increase in the demand for homes, as indicated by a 22-percent increase from 1982 to 2010 (85-percent increase through 2006) in the real home price index compiled by economist Robert Shiller;⁴¹ (2) an increase in the median square footage of floor size of new single-family homes, from 1,520 square feet in 1982 to 2,169 square feet in 2010;⁴² and (3) an increase in the homeownership rate, from 65 percent in 1982 to 69 percent in 2005, before a decline to 65 percent in 2012.⁴³ (These trends are in addition to those noted in endnote 37.)

The impact of nonshelter service categories. Among major service categories other than shelter, medical care services increased in relative importance over the study period, albeit to a lesser degree than shelter, while transportation services and energy services decreased in importance. The share of expenditures that consumers allocated to “other services” (e.g., cable and satellite television and radio service; video cassettes, discs, and other media including rental; and Internet services and electronic information providers) increased from 5.1 percent to 11.6 percent over the period.

As was the case for shelter and OER, the shift in expenditures from commodities to “other services” was not a pure price effect. The quantity of “other services” increased by 56.1 percent over the study period, while the quantity of commodities declined by 9.5 percent; when commodities are compared with “services less other services,” the quantity of commodities decreased by only 2.7 percent, while that of “services less other services” increased by 2.3 percent.⁴⁴ This shift is not unexpected given the increased availability in the marketplace, over the last quarter century, of such services as cable and satellite television, Internet services, and other digital media capabilities. However, this result contrasts with the finding that the absolute per-household quantity of “other services” declined by 9.1 percent (see table 3), indicating that even as the quantity of “other services” purchased has declined, consumer expenditures on “other services” have come at the expense of commodities, after accounting for relative price change.

Meanwhile, there has been a large relative drop in the quantity of transportation services (an experimental index was created for “services less transportation services”) and medical care services, and a slight increase in the quantity of energy services. In short, the composition of consumption has entailed an increase in the quantity of a more diverse set of marketplace offerings and a decrease in the quantity of traditional categories like transportation and medical care, and slightly more of energy services. (Table 5 summarizes these results.)

The large relative drop in medical care services comes as a surprise, but it is important to recall the distinction between total consumption expenditures and consumption attributed to out-of-pocket expenditures on medical care services. The CPI is a measure of the average change over time in the prices paid for a constant-quality market basket of goods and services. The CPI, then, is a measure of consumption derived from out-of-pocket expenditures and not necessarily a measure of total consumption. As explained earlier, it is indeed the case that total expenditures on medical care services for the benefit of consumers have increased, while consumer out-of-pocket expenditures on medical care services have decreased; as a result, the quantity of medical care services paid for out of pocket has decreased relative to that of commodities.

View Chart Data

Durables and nondurables as shares of the CPI-U for commodities, December 1982–December 2012 (in percent of commodities)

Year	Durables	Nondurables
1982	24.7	75.3
1983	25.4	74.6
1984	25.5	74.5
1985	25.0	75.0
1986	26.7	73.3
1987	26.1	73.9
1988	25.7	74.3
1989	25.0	75.0
1990	23.6	76.4
1991	23.9	76.1
1992	24.0	76.0
1993	24.3	75.7
1994	24.4	75.6
1995	24.5	75.5
1996	23.9	76.1
1997	27.2	72.8
1998	27.0	73.0
1999	25.9	74.1
2000	25.3	74.7
2001	29.3	70.7
2002	28.0	72.0
2003	28.1	71.9
2004	27.3	72.7
2005	28.4	71.6
2006	27.6	72.4
2007	26.3	73.7
2008	26.6	73.4
2009	25.1	74.9
2010	24.4	75.6
2011	22.4	77.6
2012	22.1	77.9
Source: U.S. Bureau of Labor Statistics.

Durables versus nondurables. Additional insight into long-term changes in the composition of expenditures can be obtained by disaggregating commodities into durables and nondurables. Figure 6 shows that the relative importances of durables (e.g., appliances, new vehicles, televisions, and sporting goods) and nondurables (e.g., food, apparel, medical care commodities, and motor fuel), calculated as shares of the relative importance for the CPI-U index for commodities, have remained relatively stable over the period. But a closer look at the price and quantity effects of these changes in expenditures reveals that consumers have purchased relatively more durables than nondurables, as they have shifted to more services. Thus, the general result of a shift from commodities to shelter remains unchanged. Further, regardless of whether durables and nondurables are compared with services or with “services less rent of shelter” and “rent of shelter,” consumers have increased their purchases of durable goods while reducing their purchases of nondurables.

View Chart Data

Relative change in price and quantity for durables, nondurables, and "rent of shelter," December 1982–December 2012 (in percent)

Category	Price	Quantity
Durables	-51.9	39.4
Nondurables	-5.5	-17.9
Services	19.0	7.6
Durables	-50.6	57.6
Nondurables	-3.0	-7.2
Services less rent of shelter	29.1	-2.6
Durables	-47.5	31.7
Nondurables	3.3	-22.5
Rent of shelter	24.1	28.1
Source: U.S. Bureau of Labor Statistics.

Comparing commodities with “services less rent of shelter” reveals that, from December 1982 to December 2012, the quantity of items purchased increased by 57.6 percent for commodities, decreased by 7.2 percent for nondurables, and decreased by 2.6 percent for “services less rent of shelter.” A similar comparison of commodities with “rent of shelter” shows that, over the same period, the quantity of durables increased by 31.7 percent, that of nondurables decreased by 22.5 percent, and that of “rent of shelter” increased by 28.0 percent. (See figure 7.) Similar results can be obtained when comparing durables and nondurables against OER. (See figure 8.) This finding may reflect increased purchases of durable goods, as well as shelter, after traditional necessities like food and clothing have been covered.

View Chart Data

Relative change in price and quantity for durables, nondurables, and OER, December 1982–December 2012 (in percent)

Category	Price	Quantity
Durables	-51.9	39.4
Nondurables	-5.5	-17.9
Services	19.0	7.6
Durables	-51.3	56.8
Nondurables	-4.3	-7.7
Services less OER	25.0	-1.6
Durables	-46.0	29.4
Nondurables	6.2	-23.8
OER	26.3	47.0
Source: U.S. Bureau of Labor Statistics.

None of these findings indicates that the absolute consumption of any particular category of goods and services has decreased. The analysis pertains to relative changes in quantity. The increase in the quantity of durables relative to nondurables does not mean that consumers buy less food and clothing. Durable commodities comprised only 8.8 percent of total consumer expenditures in 2012, compared with 13.1 percent in 1982; nondurable commodities comprised 31.0 percent of expenditures in 2012, compared with 39.9 percent in 1982. Consumers still spend more on nondurable items such as food and clothing than on durable items like televisions and cars, but nondurables have become relatively less important in the overall market basket of goods and services.

Clearly, then, not only has a shift in consumption occurred between commodities and shelter, but the quantity of durables has increased while the quantity of nondurables has decreased—a result that holds regardless of the overall increase in the consumption of shelter. To the extent that the decline in the relative quantity of nondurables might reflect a decline in the quantity of medical care commodities, it should be recalled that CPI expenditure data reflect out-of-pocket expenditures. A decline in the quantity of medical care commodities does not necessarily imply less consumption of medical care commodities; it implies only that an increase in consumption (as might be reflected in PCE data) is covered by employer-provided and other third-party health care insurance premium support.

A similar outcome occurs when comparing durables and nondurables with major categories of services. As shown above, from December 1982 to December 2012, the quantity of shelter, “other services,” and energy services increased relative to commodities, while consumption of transportation services and medical care services decreased. Further, it is observed that consumption increased for durable goods, decreased for nondurable goods, increased significantly for shelter and “other services,” increased less significantly for energy services, and decreased for transportation services and medical care services. (Table 6 summarizes the results.)

THE U.S. ECONOMY HAS CHANGED from being a manufacturing-based economy to being one based largely on services production. This change has been accompanied by a shift in the allocation of consumer expenditures from commodities to services, despite a relative decrease in the prices of commodities. This article has examined the price and quantity components of changes in expenditures to identify the particular categories of services driving the overall shift to services consumption. Focusing on absolute changes in per-household expenditures during the 1984–2011 period, the article found a 9.1-percent increase in the quantity of services and no change in the quantity of commodities. This trend has been driven by an 18.5-percent increase in the quantity of shelter and a 20.2-percent increase in the quantity of OER. Meanwhile, the absolute quantity of durable goods has increased, while the absolute quantity of nondurable goods has decreased. The quantity of health care services decreased by 17.8 percent, although the share of PCE accounted for by health care services increased from 4.7 percent in 1959 to 16.2 percent in 2009. This difference occurs because PCE data account for third-party expenditures, while CE and CPI data account only for out-of-pocket expenditures.

Focusing on relative expenditures for the 30-year period from December 1982 to December 2012, the analysis found an increase in the total quantity of services at the expense of commodities—a shift primarily driven by an increase in the quantity of shelter, in particular owner-occupied shelter. This shift has paralleled an increase in housing demand and homeownership rates over most of the last three decades, as well as an increase in the size of new single-family homes. The analysis also found that the quantity of energy services and “other services” has increased relative to commodities, while the quantity of transportation services and medical care services has decreased. The decrease in quantity of medical care services does not mean that consumers have consumed less medical care services; rather, it means that third-party premium support has picked up the tab.

In addition, the quantity of durable goods has increased relative to nondurable goods. This shift from nondurable to durable commodities is observed regardless of whether durables and nondurables are compared with “rent of shelter” or “services less rent of shelter.” This result also holds when comparing durables and nondurables with OER, “other services,” transportation services, medical care services, and energy services. In short, CPI and CE data show that the shift to a services-based economy entails more and bigger homes, more of “other services” (such as Internet services and cable television) relative to commodities, more durable items, and more energy services in 2012 than in 1982, while PCE data indicate that total consumption expenditures on health care and financial services have increased.

Appendix