Enugy, Vol

I, pp. N-178.

Pergamon Press 1976.

Printed in Great Britain

ENERGY

COST OF LIVING?

ROBERT HERENDEENS and JERRY TANAKA

Center for Advanced Computation, University of Illinois, Urbana, IL 61801,U.S.A.

(Received 1March 1976)

AbstractWe evaluate the energy requirements of household expenditures for all products from the 1!960-61 Consumer Expenditure Survey of the Bureau of Labor Statistics. We use more detail and employ more accurate energy intensities than in a previous, preliminary work (R. Herendeen, Affluence and energy demand, Mechan. Engng, October, 1974),and also introduce a modest analysis of errors. We find that, within error bounds, one universal curve shows the dependence of energy impact of expenditures for households of 2 through 6 members. This curve bends down somewhat; that is, it is less than linear. The single-member household falls below the universal curve, apparently because of reduced purchases of actual energy. A typical poor household exerts -65% of its energy requirements through its purchases of residential energy and auto fuel; for an affluent household, this fraction drops to 35%. We also find evidence that urban life is approximately 15% less energy intensive (Btu per dollar expended) than rural non-farm life. 1. INTRODUCTION

Only one-third of Americas energy is used in residences or the private automobile. Since private consumers are responsible for a total of three-fourths of the entire energy budget (the rest being required to support government expenditures and exports), we can say that the average household consumes more energy indirectly through the purchase of goods and services than directly though the purchase of energy itself (see Fig. 1).

Fig. 1. Role of persona1 consumption in U.S. energy demand, l%7. PCE means personal consumption expenditures. The direct component includes energy penalty on energy purchases, such as conversion losses in power plants.

In this study, we evaluate empirically the relationship between household expenditures and total resulting energy requirements (which we call energy cost), with especially detailed treatment of the non-energy purchases. Particular motivation was given by the often-quoted observation that, since household-energy purchases tend to saturate with increasing income, increased energy prices must have strongly regressive effects. On the other hand, we noted that many non-energy expenditures showed rapid increase with increasing income (housing, education, air travel): these must be energy-costed as well. To convert expenditures to energy requirement, we use energy input-output analysis.. This accounts for all energy consumed in the economy to support a certain activity, including contributions all1along the mining-manufacturing-sales chain. The most recent results divide the economy into 357 sectors, but here we aggregate to fewer sectors.
tWork supported by the National Science Foundation. SPresent address: Institute of Social Economics, Technical University of Norway, 7034 Trondheim-NTH, Norway. 165

166

R. HERENDEEN and J. TANAKA

Our source of household expenditure data is the Consumer Expenditures Survey of the U.S. Bureau of Labor Statistics (BLQ3 These are extremely detailed, cataloging yearly expenditures down to the last nickel for 13,000 U.S. households. Since every activity required energy, this comprehensiveness is necessary. The price we pay for completeness is age; the last BLS survey for which results are available was conducted in 1960-61. (A similar survey covering 1972-73 should be available in mid-1976.) The results in this study, therefore, must really be interpreted as offering a baseline. 2. ENERGY COSTING Use of input-output economics to obtain energy cost has been described before,** and applications have appeared several times.& It is an empirical approach, based on the U.S. economy for a particular year. It explicitly accounts for the implied (or embodied) energy associated with any dollar transaction, and traces every consumer product back to its basic raw materials, taking into account that different industries pay widely different prices for their fuels. One question is, of course, if the energy intensities (Btu/$) of different products are really different. We have found that they differ by as much as a factor of ten when measured at the point of manufacture (e.g. extruded aluminum vs a haircut). However, the consumer price of most goods contains a sizable wholesale and retail markup which tends to push the energy intensities towards a common value. But a significant spread still persists (Table 1). We measure energy intensity in primary terms (coal plus crude oil plus gas plus a primary equivalent of hydro and nuclear electricity.) By energy, we mean the enthalpies of combustion. We take no account of the actual specific use (feedstock, low or high temperature heat, motive power, lighting, etc.) and the implications for substitution these might have.
Table 1. Energy intensities used for the 68 consumption categories. These have been given overall capital and inflation corrections to account for the fact that Ref. [2] was the basic source of energy intensities in producers prices; Ref. [2] is based on 1%3 data. The capital factor is 1.053(multiplicative); this assumes that half of U.S. capital purchases are for maintenance and replacement. The inflation factor is 1.026 (multiplicative). Intensities given here are thus in Btu per $ 1%1. Gas rate, electric rate, and split rate indicate that actual national-average rate structures were used. The last two columns refer to error analysis. For coefficient error, 1 denotes best (that is, smallest), 2 denotes medium, and 3 denotes worst. Expenditure error is defined as P, in Appendix A, and discussed there Sector 1 2 3 4 5 6 I 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Categories Food prep. at home Food away from home Alcoholic bev. Tobacco prod. Rented dwelling, total Owned dwelling, other Owned dwelling taxes Owned dwelling repairs Owned vacation home, other Owned vacation home, taxes Owned vacation home, repairs Lodging out of home city Other real estate Water and sanitary service Coal and coke Wood Kerosene Fuel oil Other solid and petr. fuels Gas Electricity Gas and elec. combined Household oper. Laundry supplies Cleaning supplies Household paper Telephone and telegraph Household textiles Furniture Floor coverings Energy intensity (Btu per $ 1%1) 61240 50416 53807 34513 22981 11771 0 72161 11771 0 72161 48575 17957 140797 1.5889 x 1@ 1.1686 x 10 1.2522 x lo6 1.13626 x 10 1.13626x10 Gas Rate Electric Rate Split Rate 42891 Coefficient error 2 2 Expenditure error (%) 0.5 1.9 0.5 0.5 1.5 2.1 2.4 2.1 2.1 2.1 2.1 2.1 2.1 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.o 1.5 1.5 I.5 1.5 1.1 2.4 2.4 2.4

2 2 2 2 2 2 2 2 2 2 2 1 2 2 2

111530 104635 58189 71096 2 2

Energy cost of living Table 1(Contd) Energy intensity (But per .$l%l) Coefficient error Expenditure error (%) 1.8 3.4 3.9 5.0 1.3 1.8 3.0 1.5 3.0 3.0 3.0 3.0 3.0 3.0 1.3 3.0 2.8 1.5 1.4 1.4 1.1 1.1 1.0 1.0 1.6 2.1 1.7 4.7 5.0 1.3 2.2 2.0 16.0 16.0 16.0 16.0 16.0 16.0

167

sector 31 32 33 34 35 36 37 38 39 48 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68

Categories

Major appliances Small appliances Housewares Misc. household items Clothing materials and services Clothing upkeep Auto purchase Motor gasoline Motor oil Lube, washing, etc. Tires Batteries, etc. Other operating expenses Repairs and parts Auto insurance Registration and other expenses Public transp., home city car pool Public transp., out of home city Other transportation Medical care Drugs Personal care Personal care supplies Recreation Spectator admission Reading materials Education Miscellaneous Personal insurance Gifts and contributions Cash in bank Purchase of non-farm dwelling Purchase of farm dwelling Purchase of other real property Investment in business Stocks and bonds Other assets

74448 2 2

.50140 51037 47265 80019 520247 69578 38976 95655 50140 38976 36126 101271 104945 520247 141891 520247 41482 57517 50140 66082 50140 30102 55246 58046 50140 36126 0 54050 76163 82474 82474

2 2

2 2 2

3. EXPENDITURE

DATA

Besides the problem of age, the main difficulty with the BLS survey data is in matching their expenditure categories with the input-output categories. We used three methods. (1) Matching directly to input-output categories. To convert the energy intensities to purchasers (i.e. consumers) price, we used data on margins from the Bureau of Economic Analysis (BEA) of the U.S. Department of Commerce. (2) Matching to BEAs personal consumption activity categories. This is different from that in (1) above; it represents BEAs attempt to convert familiar consumer activities (e.g. a meal in a restaurant) into its component input-output expenditures, as given in their publication Personal Consumption Expenditures in the 1%3 Input-Output St~dy.~ (3) For a few sectors, use of independent data, e.g. converting expenditures for natural gas to energy using national average rate structures which explicitly include volume discount pricing. The matching problem is a limiting one; as a result, we have aggregated the BLS expenditure data into 68 categories. Table 1 lists these and the corresponding energy intensities. Documentation is given in Ref. [lo].
4. RESULTS AND ANALYSIS

The BLS survey covered some 13,000 households and included data on income, number of members, location, age of family head, etc. If we had access to the raw data, we could analyze statistically for the most significant variables. Unfortunately these detailed data are not available;

168

R. HERENDEEN and J. TANAKA

only various aggregated data are. This limits our analysis to the role of income and of location (urban vs rural) and even introduces uncertainty into these, as we will discuss later. We call attention to the fact that some of our results contain error bars. We have attempted to account for potential errors in both the energy intensities from input-output analysis, and in the expenditure data from BLS. Still, some errors had to be estimated (see Appendix A). Errors here are l-sigma; that is, the probability is about 0.69 that the true value falls within the bars.
1. Results on total household energy requirements us expenditures

Figures 2a-i show the energy requirements vs expenditures for different household size (one to six or more members). We plot total expenditures, not income after taxes, since welfare payments, etc., must be included. (The numerical data used for these curves are given in Table 2). In Fig. 2b, we plot curves for several household sizes on the same axes. From Fig. 2b, we note that the energy required by households of the same income is statistically the same, regardless of size, except for the single consumer. The single consumer does appear to use less energy per dollar, and much of the difference is due to reduced actual energy purchases. At this point, our aggregated data causes difficulty, since we can not extract, e.g. age effects to try to explain the reduced purchase. Note also that the error bars are large for single consumers because of small sample sizes. For households of 2 or more members, a fairly universal energy/expenditure curve seems to emerge. Age, location, etc. effects are buried and could be significant. For this curve (e.g. Fig. 2a), we can see that energy purchases do tend to saturate with income, while total energy requirements show a much weaker tendency to level off. Energy purchases account for roughly 2/3 of the energy impact of the lowest income households, while this ratio has dropped to almost l/3 for the most affluent. While the total energy vs expenditure curve does not saturate, it does seem to bend down somewhat, so that the average energy intensity (Btu/$) does decrease with increasing expenditures. Roughly, the average energy intensity decreases by about 30% from the poorest to the richest expenditure class. Average energy intensities are listed in Table 3. Note that these are averages ; marginal energy intensities (the slope of the curve) would show even a greater change. Whether the total energy vs expenditue curve really does bend down is controversial, so here we must refer again to the error bars and ask if it is possible statistically that it could be linear. It

I600

AVERAGE OF ALL U.S. HOUSEHOLDS

1600

. -

1400

01

25

7.5

10 EXPENDITURES

12.5

IS
1

17 5

20

225

25

( IO3 196lt

Fig. 2(a).

Energy cost of living

169

LEGEND 1600 .. . . . . ---5 MEMBERS 4 MEMBERS 2 MEMBERS I MEMBER AVERAGE OF ALL U.S

1600

HOUSEHOLDS

1400

600

400

200

01

25

75

IO EXPENDITURES

12.5

15 ( IO3 1961S )

175

20

22 5

25

Fig. 2(b).

t
SINGLE CONSUMERS

1600

H
T
600

TOTAL

/$gf ,
2.5 5 7.5 IO 12.5 EXPENDITURES

,$,
17 5

,
20

DIRECT

,
25

15

22.5

( IO3 1961$ )

Fig. 2(c).

170

R. HERENDEEN and J. TANAKA

I800

2 MEMBERS

1600

1400

600

400

200

2.5

7.5

IO EXPENDITURES

le.5

IS ( IO3 l96tS I

17.5

20

22.5

25

Fig. 2(d).

1800

4 MEMBERS

1800

1400

_ 5 0 E %i ffa E

1200

1000

800-

600

01

2.5

75

IO EXPENOtTURES

12.5 ( IO

15 l96l$ )

17.5

20

22.5

25

Energy cost of living

171

leoo

6 OR MORE

MEMBERS

1600

600

400

200

01 0

2.5

7.5

IO EXPENDITURES

12.5

15 ( IO3 196lt 1

17 5

20

22 5

25

Fig. 2(f). Fig. 2. Household energy requirements vs expenditures for different household sixes. Total refers to all purchases; direct, to residential heat and light and automobile fuel. Direct includes the energy penalty in refining, conversion to electricity, etc.

Table 2. Frequency distribution of energy intensities for 68 expenditure categories. Three categories use a rate structure, and therefore intensity depends on amount purchased Number of expenditure categories 3 3

Energy intensity (Btu/$l%l) 09999 10000-19999 2oooo-29999 3oooo-39999 4oooO-4%% 5otJcxLs9999 6oooo-6%% 7ooc4l-7!9999 8oooo-89999 !Kttw-99999 1~119999 12OOW139999 lWOO-19999 1-199999 2tXxQM99999 5ooooo-g%%g loooooo-1499999 15m

1
6 4 16 3 7 4

1
7 2 3 4

1 w

is not possible to draw a straight line through all of the error bars and it is especially hard if one also requires that the line go through the origin. In our opinion, the relationship is not linear. All this points to the conclusion that rising energy prices will have (and have had!) a regressive effect. It would not be as regressive as one would conclude from only direct energy purchases per se. To say exactly how much, one would need to know how price increases would be passed on. This should not be interpreted as meaning that the rich would be as hard hit as the poor, of

172

R. HERENDEEN and J. TANAKA Table 3. Energy and expenditure data for several household sizes, and sample error estimates for average of all households (a) Household size-average of all U.S. households; location-urban and rural combined. Income before taxes (S) 573 1548 2618 3746 4923 6044 7500 9716 13583 27750 Average energy intensity (Btu/$) 111000 105wO 102000 99900 1OlOlm 102000 98600 96300 91200 79700

Expenditures (S) 1363 1938 2958 4085 5327 6315 7641 9383 12758 23056

Energy (106Btu) 151 204 302 408 536 643 753 903 1160 1840

(b) Error calculation for Table 3(a). CoetBcient error (0s) a1 a2 ff3 Expenditures ($) 1363 1938 2958 4085 5327 6315 7641 9383 12758 23056

-o0.05 0.10 0.15 0.10 0.20 0.30 0.20 0.30 0.50

-non-zero0 0.05 0.10 0.15 0 0.10 0.20 0.30 0 0.20 0.30 0.50

3.5 2.0 1.8 1.8 1.7 1.8 1.8 1.7 1.7 2.2

Per cent error in energy 5.7 9.1 8.5 9.2 10.2 12.5 3.9 5.9 1.6 2.6 4.2 6.1 3.6 5.4 1.0 2.1 3.7 5.5 3.5 5.3 1.1 2.1 3.7 5.4 3.5 5.2 1.9 2.6 4.0 5.5 3.5 5.3 2.4 3.0 4.3 5.8 3.5 5.3 2.5 3.0 4.3 5.9 3.5 5.2 2.3 2.9 4.2 5.7 3.3 5.0 2.8 3.3 4.4 5.8 3.7 5.9 8.2 8.5 9.0 10.1

(c) Household size-single consumers; location-urban and rural. Income before taxes (S) 659 1493 2686 3907 5204 6307 7939 9908 14494 40461 Average energy intensity (Btu/%) 111000 99700 84800 79300 76900 73600 69500 69900 78800 56100

Expenditures (S) 1093 1702 2779 3716 4572 5781 7136 8988 16588 25727

Energy (106Btu) 121 170 236 295 351 425 4% 628 1310 1440

(d) Household size-2 members; location-urban and rural. Income before taxes (S)
644

Expenditures (S) 1638 2127 2900 4030 5029 6024 7217 9138 13114 23512

Energy (106 Btu) 194 239 317 423 519 609 687 820 1080 1910

Average energy intensity (BtulO) 119Ow 112000 109000 105000 103000 101000 95200 89700 82100 81300

1573 2601 3755 4993 6219 7722 9895 13986 2%10

Energy cost of living Table 3. (Contd) (e) Household size-4 members; location-urban and rural. Income before taxes (8) - 403 1623 2611 3691 4848 5985 7434 9715 13808 27604 Expenditures (8) 2620 2338 3502 4709 5866 6710 8032 12526 22894 Energy (106Btu) 340 250 386 498 645 699 807 933 1160 1840 Average energy intensity tBtu/$) 130000 107ooo 110000 110008 104000 101000 99300 92700 80400

173

(f) Household size-6 or more members; location-urban and rural. Income before taxes 6) 459 1636 2511 3549 4591 5723 7046 9192 12822 27521 Average energy intensity (Btu/$) 105ooo 107000 104Ol30 106OCQ 99800 112000 111000 107000 101000 86800

Expenditures 6) 2521 2103 3304 4399 6097 6337 7482 9194 12048 20218

Energy (lo Btu) 264 225 344 465 609 710 828 1040 1220 1750

course. The rich would change their vacation plans while the poor would go short on home heating fuel. If present cross-sectional data can be used as a basis, we also conclude that redistributing buying power away from the rich towards the poor will increase the energy intensity of personal consumption expenditures. This result is not as dire as it sounds, however, since the redistribution will result in fewer poor people; as they become richer, their energy intensity changes. For example, if all households had expenditures equal to the average in Fig. 2a, their energy requirements would be only 4.7% higher than the distribution shown in Fig. 2a (see Appendix B for details of calculation). This is, of course, a steady-state calculation, and ignores transients. 2. Details of expenditure patterns with income In Fig. 3, we break down the expenditures of three income classes (poor, average, rich) into 11 categories for a household of 4 members. We note that, in going from poor to rich, the relative portion of total energy requirements due to energy purchases decreases from 65 to 35% and that the increase of non-energy purchases is dominated by growth in travel, education, housing and investments (investments must also be energy-costed; see Appendix C). 3. Efect of urbanization on energy intensity BLSs aggregation of data again causes a problem. The average urban household differs in total expenditures from the average rural non-farm household, which complicates comparison. In Fig. 4, however, where we plot location data next to the average energy-expenditure curve, a clear trend emerges: the urban household tends to the low energy intensity side of the curve, while the rural farm and non-farm (especially the latter) tend to the high energy intensity side of the curve. We attempt to quantify this result in Table 4. We see that the urban household averages about 15% less energy intensity than the rural non-farm household and is also less energy intensive than the rural farm household; from the error bars in Fig. 4, we conclude that

174

R.

HERENDEEN. and

J.

TANAKA

.._ RECR&TlON 1.39% TRANSP.EESIDES AUT? \ 7 XJCATION 0.59% , INVEST.. INSUR. 0 PURCH. 8 MAINT MEDICAL, PERS. CLOTHING

(b)

TRANSP. BESIDES AUTO

MEDICAL,

PERS.CARE

Fig. 3. Breakdown of energy requirements into 11 expenditure categories, for poor, average and rich households with four members. (a) Expenditures = $2,620; (b) expenditures = $5,866; (c) expenditures = $22,894.

Energy cost of living

900 INSIDE 800 I 2 3 SMSA : NONFARM NONFARM FARM

URBAN RURAL RURAL

700

OUTSIDE SMSA : 4 URBAN NONFARM 5 RURAL NONFARM 6 RURAL FARM

; k8 w 0 = & 5 = : 5 *

600

500-

400

/
300 -

ALL

URBAN AND RURAL FAMILIES AN0 SINGLE CONSUMERS URBAN AND RURAL SINGLE CONSUMERS URBAN AN0 RURAL OF 2 OR MORE FAMILIES

200

A ALL 8 ALL

I
IO0 AVERAQE OF ALL

1
HOUSEHOLDS

0'

IO

2.0

3.0 ANNUAL

4.0 EXPENDITURES

50 ( t IO3 1

60

70

6.0

Fig. 4. Effect of urbanization on household energy intensity. Data points are for national average size household, by location. Direct comparison is difficult because of different average expenditures for, say, urban and rural farm households. Energy intensity, however, can be compared with the average energy/expenditure curve a point lies on. The dotted line is the average curve from Fig. 2a. A point above the curve is more energy intensive; a point below is less energy intensive. Energy intensity is expressed in Btu per dollar. Table 4. Comparison of energy intensity of urban and rural households. Listed are the percentage deviations from the average (Fig. 2a) for that expenditure Inside SMSA Urban All Rural non-farm households I Rural farm Urban Single Rural non-farm consumers t Rural farm
-1

Outside SMSA +2 +20 +11 -II +19 +5

+10 t16 -26 +28* -5*

*Indicates that error bars are too large for statistical significance.

this observation is statistically significant. We are more suspicious of the rural farm result because farms typically have trouble dividing energy bills into farm and non-farm use. Another problem in general is the definition of urban and rural non-farm households. The definitions are given in Ref. [12], pp. 7-15, and it appears that urban includes some households one would call suburban. Why is the urban household less energy intensive? In Table 5, we break expenditures into 11 categories. The urban household spends at least 20% less of its dollar budget on residential energy and automobile fuel than the rural. The resulting reduced energy use is not counteracted by the fact that the urban house hold spends at least twice as much for transportation other than auto. These data seem to confirm the picture of the urban family as one in which people live in a less-energy demanding dwelling (apartment?) and drive less than the folks in the suburbs and country.
4. CONCLUDING REMARKS

We have been able to draw conclusions about total energy cost because of our inclusion of the energy cost of all goods and services. But we repeat that the data are old and aggregated, and that
EGY Vol. I, No. 2-E

176

R.

HERENDEEN and

J. TANAKA

Table 5. Detailed energy and expenditure data for urban and rural households (averaged over all household sizes). Listed are the percentages of total household requirement. Figures with parentheses are expenditures in dollars; figures without are the resulting energy Household type, inside SMSA Rural Urban non-farm Rural farm Residential energy Auto fuel Auto purchase, maint. Transp. besides auto Food Housing Clothing Medical, personal care Education Recreation Savings, investments, insurance, Misc. 29.1 15.1 6.5 3.5 14.2 15.5 4.7 4.1 1.1 1.7 4.6 ----loo.1 (3.7) (2.7) (8.3) (1.5) (23.8) (28.2) (8.9) (8.0) (1.8) (3.5) 34.2 (4.7) 17.3 (3.6) 5.7 (8.6) 2.6 (0.8) 11.2 (21.8) 17.1 (32.7) 3.6 (7.8) 3.4 (7.7) 0.7 (1.4) 1.4 (3.3) (7.5) (99.9) 43.3 15.8 5.8 0.8 10.1 7.8 3.5 4.0 0.6 1.0 (5.6) (3.7) (9.8) (0.3) (21.5) (17.8) (8.5) (10.4) (1.3) (2.5)

(9.8) 2.7 (100.2) 99.9

7.4 (18.7) 100.1 (100.1)

Relative energy intensity with respect to average (from Table 4)

0.93

1.10

1.16

possible specific technological changes in the efficiency of use of energy in the society are simply not part of our model. Neither are changes in energy use due to increasing prices. Our intent has been to allocate energy requirements to final consumers. Many arbitrary decisions are inherent in our approach. One example is business travel. We considered this to be a part of industry and commerce and hence allocated it to final consumer products. On the other hand, it does seem to be a close antecedent of high income; indeed, it is often considered a fringe benefit of a high-paying job. We would also speculate that a good portion of it is actually discretionary. And the great majority of it is by plane, the most energy intensive mode. It can easily exceed personal discretionary travel. An informal survey of five Ph.D-holding energy researchers at the University of Illinois showed that, on the average in 1974, each one traveled 8000 miles for personal use (i.e. paid for with personal funds), of which most was by car and none was by plane. In contrast, each one traveled 13,000 miles for business purposes (i.e. paid for by someone else) and 92% was by plane.
Table 6. Per-household business and convention travel, 1972. Conversion to energy at 11,200 Btu/passenger mile, which assumes air travel and includes the indirect energy requirements as well as the airplane fuel. Passenger mile data are from the 1972Census of Transportation, Vol. I, p. 20. They were converted to a per-household basis by using data from Statistical Abstracts of the United States, 1972 Edn; p. 322, 1973Edn; pp. 40,320,322 Business passenger miles 325 620 775 1378 1860 Energy (lo Btu) 3.6 7.0 8.7 15.4 20.8

Household income ($) Less than 5,ooO 5001-7500 7501-10,000 10,001-15,ooo More than 15,000

The general trend in business travel with income is indicated in Table 6; the dependence is steeper than linear. Today, the average per household is still not that large; it amounts to 21 million Btu per year for a relatively affluent family vs about 1700 million already allocated. But the trend is insidious, as evidenced by the apparent fact that energy researchers travel much more than average. This is just one example of a possible, different allocation scheme.

Energy cost of living REFERENCES

117

I. C. Bullard and R. Herendeen, Proc. IEEE, Vol. 63, No. 3 (March 1975). 2. R. Herendeen and C. W. Bullard, CAC Document No. 140, Center for Advanced Computation, University of Illinois at Urbana. Urbana, II. 61801 (Nov. 1974); A shortened version is C. W. Bullard and R. Herendeen, Energy Policy 3(4) (Dec.

1975). 3. ConsumerExpenditures
4. 5. 6. 7. 8. 9. 10.

I I.
I?. 13.

and Income, Suppl. 3, Part A to EL.9 Report No. 237-93, U.S. Department of Labor, Bureau of Labor Statistics, (May 1966). E. Hirst, Science 184, 134 (1974). R. Bezdek and B. Hannon, Science 185, 669 (1974). B. Hannon, Science 189, 95 (1975). R. Williams (Ed.), The Energy Conservation Papers. Ballinger, Cambridge, Mass. (1975). Input-Output Structure offhe U.S. Economy, 1%7. Data tape available from US. Department of Commerce, Bureau of Economic Analysis. The tape contains margin data unavailable in published versions. Personal Consumption Expenditures in the 1963 Input-Oufpuf Study. Survey of Current Business, Jan. 1971, p. 31. This is at the 83Sector level. A similar study at the 368 level (unpublished) is available from BEA (see Ref. (8]). J. Tanaka. Formulation of computer programs for the energy cost of living model, CAC Tech. Memo. 5X. Center for Advanced Computation, University of Illinois, Urbana, IL. 61801 (July 1975). Eva Jacobs (Telephone conversations). U.S. Department of Labor, Bureau of Labor Statistics. Consumer Expenditures and Income: Survey Guidelines Bulletin 1684. U.S. Department of Labor, Bureau of Labor Statistics, (G.P.O. Document 2901-0652) (1971). Consumer Ewing Indicators. Current Population Reports Series P-65, No. 33, U.S. Department of Commerce, Bureau of the Census. 16 Oct. (1970).

APPENDIX
Error analysis

All energies are obtained from a sum of products of energy intensities times expenditurest:
E = 2 CY,.

I I

We assume the errors in the Cs and the Ys are independent;

then

!@=A
E

C(A(:Y,)+~(~,AY,))
c C,Y#

where AC, and AY, are the respective errors. We express the errors as fractions;

viz. AC, = a,C,. AU, = p,Y,; then

J(c AE _=A
E

c~Y.(a.~+B:))
c C<Y,

The as must be obtained from some estimate of the accuracy

of the entire input-output

necessarily crude. We classified all the Cs into 3 categories assumptions for the as (see Table I for the classification).

(best accuracy,

energy technique. This is medium, worst), and then tried several

The ps are meant to reflect sampling error and therefore should have some sort of inverse relationship to the actual number of households in that income/demographic class. Luckily, we were able to tind some sampling errors (f-sigma) derived by BLS for its urban (only) I%&61 consumer survey (Ref. 12, Table B-l I). These could be matched quite well with our 68 expenditure categories, as listed in Table 1. We then assumed that the error should be related to BLSs errors by the inverse square root relationship, i.e.

where P, is the error given by BLS, N, is the number of respondents in their urban survey, and N the number of respondents in the class in question. The l/Q/N can be justified on standard statistical grounds, and also is used by the Census Bureau (Ref. [13], Table C). This scheme can be expected to give large errors for the lowest and highest expenditure classes. which contain the smallest number of respondents. On the one hand, this error analysis is quite ad hoc and rough. On the other hand. the input errors are given rather conservative (large) values, so we think it is a reasonable attempt. Table 3b fists errors under several assumptions on coefficient errors. For our graphs and results, we used an intermediate assumption for the as: (best, medium, worst) = (0.10, 0.20. 0.30). Table 7 shows the relative contributions of the errors in coefficients and in expenditures. The average expenditure is $348.97~ 109/55.307x IO= $6309.69. From Fig. 2a. the energy requirements of an average

tExcept

where a rate structure

is used, as for electricity

and gas. in which case we just assume a fractional

errorequal

to (I,.

178

R.

HERENDEEN

and J. TANAKA

Table 7. Relative contributions of errors in energy intensity (o) and expenditures @I).Errors shown here apply to all U.S. families and single consumers, urban and rural, from Table 3 Percent error in energy From From energy intensity expenditures combined 5.7 3.9 3.6 3.5 3.5 3.5 3.5 3.5 3.3 3.7 8.5 1.6 1.0 1.1 1.9 2.4 2.5 2.3 2.8 8.2 10.2 4.2 3.7 3.7 4.0 4.3 4.3 4.2 4.4 9.0

Income class

APPENDIX Calculation of e#ect of income redistribution Expenditures (8) 1363.42 1937.79 2957.79 4084.95 5327.40 6314.91 7640.60 9382.71 12752.45 23055.79 Energy (lo Btu) 1Sl 204 302 408 536 643 753 903 1160 1840 No. Householdsm (W 2.052 5.628 6.112 6.530 7.338 7.012 8.352 7.423 3.742 1.118 55.307

Energy For This Class (IO=Btu) 309.9 1148.1 1845.8 2664.2 3933.2 4508.7 6289.1 6703.0 4340.7 2057.1 33799.8

Total Expenditure For This Class (8 107 2.98 10.91 18.08 26.67 39.09 44.28 63.81 69.65 47.72 25.78 348.97

(a) from Table 3; (b) from Ref. [3]. household with this expenditure is 640x lo Btu. The total if all were average is 55.307x 10x 640x lo6 = 35.40x lOI Btu which is 4.7% greater than 33.80x 10 Btu, the total given above, without redistribution.

APPENDIX
Investment-energy intensity

The energy coefficients used here are based on the assumption of steady state. Capital expenditures by industry needed for replacement have been accounted for, but not those for growth (we estimate the latter to require about 5% of the nations energy in 1%3). People who invest are thus stimulating growth capital purchases; a dollar lent to a bank or spent for stock soon finds its way into a construction project, new business, and so on. The average energy intensity of all capital purchases in 1963was 71,300Btu/$. There is some dilution, undoubtedly, so we use 50,CKlO Btu/$ for the energy intensity of investment, and assign to it our worst error assumption, 230%. This value is changed slightly by additional capital and inflation corrections; see Table 1).