Digitized by the Internet Archive

in 2011 with funding from

University of Illinois Urbana-Champaign

http://www.archive.org/details/geographicalcost1491mcma

FACULTY WORKING PAPER NO. 1491

Geographical Cost of Living Differences: An Update

Walter W. McMahon

ocj

U***>0r

*>*

■t /.

9BB

College of Commerce and Business Administration Bureau of Economic and Business Research University of Illinois, Urbana-Champaign

BEBR

FACULTY WORKING PAPER NO. 1491

College of Commerce and Business Administration

University of Illinois at Urbana- Champaign

September 1988

Geographical Cost of Living Differences: An Update

Walter W. McMahon, Professor Department of Economics

Abstract

Geographical Cost of Living Differences An Update

This paper develops a method for estimating current differences in the cost of living among states, as well as differences among counties within states. It tests hypotheses relating to the determinants of these differences based on a newly refined theoretical framework, orig- inally developed by the author. It finds the key determinants to be differences in housing costs, and demand-side-related differences in per capita income. Population change cuts both ways, with upward effects on the cost of living more than neutralized by the attraction of industry to lower cost areas and by other factors.

New estimates are presented for the differences in the cost of living among states and within states for 1988, along with the result- ing production equation that can be used for later years and within other states. Direct collection of price and budget study data within all of these areas would be prohibitively expensive. Large differences in living costs emerge; higher in the East and lower in the South and rural areas with shifts since 1977 positively related to the differ- ences in economic growth rates since that time.

Geographical Cost of Living Differences: An Update

Walter W. McMahon*

There are significant differences in the cost of living among dif- ferent parts of the country, as well as among different rural and urban counties within the same state. But there are no systematic re- ports of these differences by state or by county of the type presented in this paper.

A systematic procedure for estimating these differences based on the Bureau of Labor Statistics data for selected localities was developed earlier by McMahon and Melton (19 78). The resulting esti- mates found many uses, but the estimates were for 1977. Since then an oil price shock occurred in 1979 affecting oil producing and oil con- suming states differently, followed by a major 1980-84 recession with larger effects in industrial states and a high priced dollar that cur- tailed farm exports. All of these could be expected to lead to dif- ferential effects on prices and a changed pattern of geographical cost of living differences.

The ideal way to evaluate these differences would be to collect price data from each county in every state, and to also conduct detailed budget studies of family expenditures in each county in the nation to establish the necessary weights. This procedure would be prohibitively expensive, however, and therefore likely will never be done. The Bureau of Labor Statistics, furthermore, discontinued col- lecting and publishing its cost of living index for selected locali- ties in 1981. It was this cost of living as measured by standard

-2-

budgets for a given standard of living for a typical family of four that was the basis for the McMahon-Melton analysis and estimates for the nonsampled areas.

This paper will update the procedure, as well as the estimates, adapting the new method used to this reduced data availability. The resulting new estimates for 1988 of differences in the cost of living among the 50 states, and among counties within one state (Illinois), then will be presented. The paper concludes with a brief analysis of the nature of changes in the geographical differences in the cost of living between 1977, the date of the earlier study, and the present.

I. Existing Cost of Living Measures and Their Uses There currently are no measures of differences in the cost of living among states or any other areas since the discontinuation of the BLS standard budget series for 23 localities in 1981. A Consumer Price Index (CPI) series continues to be published for the four major regions, including urban and rural breakdowns within regions plus the CPI's for 15 major cities, as shown in Appendix A. But these are not available by state, or by county. They also do not show inter-area differences in living costs, because the geographical CPI takes all budgets in the base year as the same (1982-84 = 100), whereas in fact the cost of living in these different places in the base year differs considerably.

The method adopted therefore seeks to take these base-year differ- ences in the cost of living into account by using the last report for a family cost of living budget reported by the BLS (1982) for the Fall

-3-

of 1981. This is updated to March of 1988 using the changes in the Consumer Price Index, which thereby takes both the changes in prices and the differences in the base year cost of living into account. However the Consumer Price Index also does not apply to states, but instead to the urban and rural areas within geographical regions and to a few big cities. So to relate to this, the population living in the urban Standard Metropolitan Statistical Areas (SMSA's) as a per- cent of the population in the non-SMA areas taken from the Census, U.S. Department of Commerce (1980) was used to get a weighted average of the urban and rural components of the Consumer Price Index. The resulting adjusted cost of living index for the BLS' 23 different localities then becomes the dependent variable used in the regression analysis. The logic of the model, and how each of the three explora- tory variables chosen can be used to predict the cost of living index for the other states and for the counties within states is developed below in Section II.

To consider the concept of a cost of living index, geographical differences in the cost of living affect the purchasing power of wages and salaries, which are always paid in nominal dollars, at different locations. For salaries to be comparable in real terras they therefore must be deflated (i.e., divided by) a geographical cost of living index such as the one developed here. To avoid questions of interpersonal comparisons of utility, the BLS' concept of a standard budget for a family of four, which we use here, is one that seeks to keep the head of the household on the same indifference curve with respect to com- modities purchased irrespective of where he or she locates.

-£-

This concept does not include special non-monetary returns (e.g., sunshine or seaside locations) or benefits that can sometimes partly justify the higher costs and that also affect location decisions. It is limited to differences in the monetary costs of living such as dif- ferences for comparable housing accommodations in different places, which can be substantial.

The uses that have developed for geographical cost of living indices, as well as an interpretation of its misuses, depend upon this concept. It is useful to employees in making decisions to locate because, to the extent that the cost side is to be considered in making these decisions, it is what the salary will buy in real terras, not in nominal terms, plus their evaluation of the non-monetary returns that basically govern the outcomes. That is, the evidence is strong that employees tend to make a correction for price level and cost of living differences, as well as non-monetary benefits, albeit impli- citly, and that there is no substantial money illusion (after allowing for lags in adjustment). Because of this behavior, multiplant firms with plants in different locations, state school systems with urban and rural unit districts, universities competing in inter-state job markets, and other kinds of employers who wish to maintain salaries that are comparable in different locations (plus or minus the non- monetary environmental fringes) must also normally make some adjust- ment either explicitly or implicitly for the more purely nominal differences .

A geographical cost of living index is not the same as an educa- tional price index, however, since it does not include an index of the

-5-

price of all the things with appropriate weights that a school dis- trict or university purchases. Nevertheless it is sometimes used as a proxy. The State of Florida for example has used the Florida Price Level Index, which is an index of living costs in Florida counties, as an adjustment factor in its school aid formula. Similarly, analyses of the adequacy of the resources provided for education, including interstate comparisons such as the recent study by A. Hickrod et. al. (1987, p. 9) often seek to rerrcve some of the nominal differences in costs in this way. There are non-monetary differences in benefits that probably justify only part of the cost differences among dif- ferent localities. The justification for making such an adjustment is that teachers migrate from district to district depending on the real, and not the nominal, salary. This real salary (i.e., after adjustment by a geographical cost of living index) therefore serves as a proxy for the supply price for teachers with a given level of training, ability, and experience, and hence for a given quality of education provided by those teachers, especially since salaries account for about 80 percent of most education budgets. Geographical differences in prices for items in the other 20 percent of the budget reasonably can be expected to be highly correlated with the same geographical differences in the cost of living that affect real salaries (e.g., housing and construction costs), even though the correlation is not perfect .

However the non-monetary attractions or detractions of the job also need to be factored in to get a true real supply price. As pointed out by Barro (1981, p. 7) there are many -factors that make a

-6-

school district more or less attractive to professional staff other than differences in nominal salary and the local cost of living. These other factors also influence the supply price of staff to the district. For example, a further addition needs to be made to a nomi- nal salary to compensate for the student population in especially unattractive neighborhoods. One of the more complex approaches is to develop separate simultaneous demand and supply equations for deter- mination of teachers' salaries at the district level, and then after controlling for the average level of teachers experience, remove the demand-side influences on salary (such as income, property value, and local "tastes" for education) to isolate the supply-side effects on the supply price. This simultaneous equation approach is used by Brazer and Andersen (1975), Boardman, Darling-Hammond, and Mullin (1979), Wentzler (1979), and Loatman (1980).

Although the main uses of geographical cost of living indices by employers and employees that were mentioned first are more direct, there has been continuing interest in these simple purely supply-side related indices for use in school aid formulas. For this purpose a cost of living index has the disadvantage of not reflecting all of the influences on the supply price of teachers. But it does not have the disadvantage that plagues all of the other cost of education indices that start with data on teachers salaries and use complex methods (or sweeping assumptions in the case of the hedonic price index approach) to remove demand side influences. The cost of living applies to everybody in the locality, not just teachers who are a very small fraction of the total population in the locality, and therefore from

-7-

the point of view of the school district is completely supply-side oriented. It also has the merit of simplicity. If taken as a first approximation that omits differences in the nonmonetary advantages or disadvantages of the environment of the school, its use may be better than making no adjustment at all to nominal values.

II. The Theory and The Model There have been previous attempts to investigate the sources of differences in the cost of living. Sherwood (1975), for example, used the BLS indices and price data to construct standard budgets that iso- late the effect of climatic differences on costs. But his indices are limited to this one source of differences and also were constructed for only the 44 cities and regions in this BLS sample. Haworth, Rasraussen, and Mattila (1973) and Alonso and Fajans (1970) explored the extent to which urban population and other variables explain dif- ferences in the cost of living within the BLS sample. But they did not undertake predictions for nonsampled areas. Alonso (1970) finds urban population size, when income is included, to be of minor sig- nificance. Israeli (1977) found that housing differences were a good predictor of the differential in nominal wages and prices among selected cities. But the only major efforts to extend cost of living indices from sampled to nonsampled areas have been by Simmons (1973, 1988) and by McMahon and Melton (19 78). Simmons sampled prices in 12 Florida counties and then used regression equations to extend these prices to all counties in the state. The first result, in the absence of budget studies to obtain the necessary weights, is therefore closer

-8-

to a geographic price index than to a cost of living index. Augmented by budget studies, it has been used by the State of Florida since 1978 in the Florida school aid formula. But the expense of collecting the price data, doing the consumer expenditure budget studies, and con- structing and updating the index limits the extent to which it can be extended to other states. McMahon and Melton (1978) developed a model that explains cost of living differences within the BLS sample, and then used the regression coefficients, together with measures of the explanatory variables for the non-sampled areas, to predict the cost of living index for all 50 states and for counties within California, Illinois, Pennsylvania, and Texas. But as indicated above, the index was for 1977, the data availability has changed, and there is need to update that index.

Economic theory suggests that changes in the effective demand for goods and for housing, especially when supplies are not perfectly elastic, can play a large part in the determination of geographical differences in living costs. As effective demand rises, the prices of land especially and any other goods for which supplies are not easily transportable and are therefore less than perfectly elastic rise, causing living costs to increase.

The demand function for any given locality shown in Equation (1) below expresses the quantity demanded primarily as a negative function of price (a < 0) , a positive function of per capita income in the locality (a„ > 0) and a positive function of the stock of consumption habits and/or assets is measured by the price (or value) of housing (o3 > 0):

-9-

(1) q - a p + a2Y + a3V + o^AP + \i±

Here p = a price index relevant to goods and services purchased in the area, q = a market basket of goods and services needed to sustain a family of four at a given level, irrespective of the area, [c = pq = the cost of living] ,

Y = per capita income in the locality,

V = value of the house of given size and quality (measured here

as the median value of a house available from Census data), AP = percent change in the population in the area, from 1980 to

the present, and u = disturbances. The factors shifting the demand function, Y, V, and AP, can first be considered briefly. Individual income is a critical element in the demand for virtually all goods and services, raising demand since most goods are normal goods (ot > 0) when income is higher, and where supply is inelastic (as in the case of land prices), more or less per- manently bidding up the price.

Consumer demand is also affected by a stock effect, reflecting assets and/or a stock of past consumption habits, measured here by V, the value of the housing. This stock-habit effect is sometimes measured by using past consumption as a proxy, which is tantamount to permanent income or permanent wealth by means of a Koyck transfor- mation. The Life Cycle Hypothesis of Ando and Modigliani (1963) measures it by using the total stock of assets or net worth. But such

-10-

a comprehensive measure of all assets is less relevant for purposes of analysis of geographical price differences than are the assets in the locality in the form of housing. Sherwood (1975, p. 14) found that housing costs vary widely among areas, ranging from an index of 168 in Boston to 68 in Austin, Texas. It is not only that land is immobile resulting in an inelastic supply, so that when demand rises, housing prices are driven up more or less permanently. But it is also that climatic differences have long run effects on differences in housing costs. Additionally, imperfect competition in the construction trades and building materials industry contributes to the inflexibility of prices. Using the value of the median house in a locality as a measure of past asset accumulation (and consumption habits) has the further merit of being a measure that is widely available for locali- ties from the Housing Census, whereas the less relevant more compre- hensive asset measures are not.

Population growth can have ambiguous effects on prices, as was stressed earlier by McMahon and Melton (1978, p. 326). Rapid popula- tion growth can increase the pressure on some facilities other than housing, and act to raise their prices (a > 0). On the other hand, economies of scale in certain services such as schools also can be achieved as pointed out by Alonso (1970, pp. 72-75), (a, < 0). Fur- thermore, as population migrates toward lower cost areas as it did in the early 1980s to Texas, Georgia, Kentucky, and Colorado, for example, the correlation between the population increase and the geo- graphical price index would be negative (a < 0). The net effect can- not be inferred from economic theory, but because of the large

-11-

migration toward the south and the sun belt states that occurred since the 1980 Census, it is postulated that this relationship will be nega- tive (a, < 0).

The supply equation expresses price as a positive function of the quantity supplied both in the short run and in the long run (a > 0), as well as of housing costs (a > 0):

(2) p = a5q + a6V + u2

where M9 = disturbances, and all other variables have been defined under Equation (1). Assuming linearity, the demand and supply func- tions may be solved simultaneously eliminating q. The resulting reduced form price equation then can be multiplied throughout by the appropriate quantity weight q representing the market basket of commo- dities in the standard budget for a family of four. Since these quan- tity weights are designed to maintain the same level of well being in each area, they are treated as constants and as part of the parameters in Equation (3) below. This result contains the key determinants of the cost of living, C, in each locality:

_ a q (a +a ,/a -)q a,q

(3) C = pq = -r-r^ Y + f^ — V + —^ AP + V.

*H 1/c^-o^ l/a5-a1 l/a5-a1 3

Since a < 0, the denominators can be expected to be positive. The first two numerators can be expected to be positive as suggested above, and the sign of the third numerator is indeterminate.

-12-

III . Estimation of the Model The parameters can be simplified as shown in Equation (4), the model to be estimated. Here 8 and 8 are expected to be positive, and 8^ is indeterminate, but probably negative:

(4) C - BY + 82V + 83AP + v

The definitions and data sources for the variables are:

C = Cost of Living Index for the 23 SMSA's published by the U.S. Bureau of Labor Statistics (1982.6, p. 45). These are updated to apply to March 1988 by use of the Consumer Price Index from the U.S. BLS (1988.6, p. 97) shown in Appendix A. A weighted average of the urban and rural components of the CPI in each region was used, with weights consisting of the percent of the population that is urban vs. rural in each state from the U.S. Bureau of the Census.

Y = Per Capita Personal Income, in thousands of dollars. For

states this is for 1987-IV from U.S. Department of Commerce (1987.4, pp. 72-3), and for counties in Illinois it is for 1986 from (ibid. pp. 56-7) as shown in Appendix B.

V = Value of a Standard House; measured as the median value of a

house for 1980, the latest year available, from the Census of Housing, U.S. Department of Commerce (1980, HC80-1-A). AP = Percent Change in Population, from 1980 through 1987, from

Current Population Reports, U.S. Department of Commerce (1988, p. 16, Table 1).

-13-

The results obtained for the regression which together with the data are shown in more detail in Appendix C is as follows. The t-statistics are shown below in parentheses:

(5) C = 56.66 + 3.69Y + .292V - .689AP R2 = .709

(4.25) (4.16) (2.71) (-2.75) F = 15.43 Prob. F - .0001

DW = 2.09

The signs are as expected and the t-statistics indicate that all coef- ficients reach a high .01 level of significance or above. Multi-

collinearity among the explanatory variables is sufficiently low

2 (under .47 as shown in Appendix C and the R as shown above is quite

good for cross section data. The sample is too small to partition it

into four subsets and use seemingly unrelated regressions. But the

alternative procedure used of weighting the urban and rural indices by

that state's urban vs. rural population distribution is more precise,

and therefore is a superior procedure to using seemingly unrelated

regression methods or regional dummies. It also relates somewhat more

precisely to rural school cost and consolidation issues, such as those

considered by Ward (1988, pp. 4-5).

Other regressions were tested, using population levels in place of

the change in the popualation over time for example. The Consumer

Price Index which is a major component of cost of living differences

was also explored as a dependent variable. But it has the disadvantage

of being independent of differences in the cost levels in the base

year. However none of these steps significantly improved upon the

result shown in Equation (5).

-14-

Differences in the cost of housing still emerge as by far the most significant source of differences in the cost of living. They account for about 23 percent of a typical household budget. Higher per capita incomes also account for some of the difference, especially in Connecticut and the Northeast. The effect of the growth of population is not a major factor, consistent with Alonso's (1970) earlier results. It is almost swamped, in fact, by the more recent tendencies in the U.S. for some industries and population to gravitate toward the lower cost of living in the new South and the more recently developing areas .

IV. Geographical Differences in the Cost of Living

The Results

By States. The differences in the cost of living among the 50 states and the District of Columbia are shown in Table 1. They are obtained using the regression equation (5) together with measures of per capita income (1980), value of a standard house, 1980, and percent change in the population from 1980 through 1987 measured for each state as shown in Appendix D. The cost of living index then was normalized so that 100 represents the national average for all states weighted by their population.

These results indicate that there is a 53 percent variation in the cost of living among states. The higher cost of living states continue to be in the East, Connecticut (123.7), New Jersey (119.1), and the District of Columbia (124.9) in particular plus Hawaii (113.9). In these places higher incomes and higher housing costs are both a factor. The lower living cost states are those in the South,

-15-

Table 1

Differences in the Cost of Living Among States, 1988

Percentag

e

Percentage

Index

Change

Index

Change

State

1988

1977-88

1988

19 7 7-88

Alabama

86.9

-0.3

Montana

91.6

-5.3

Alaska

101.7

n.a .

Nebraska

100.3

5.2

Arkansas

84.8

-0.9

Nevada

97.1

-9.1

Arizona

88.0

-11.3

New Hampshire

101.9

-4.4

California

110.2

2.2

New Jersey

119.1

2.1

Colorado

101.6

1.0

New Mexico

83.6

-12.1

Connecticut

123.7

2.9

New York

110.7

0.3

Delaware

101.7

-8.5

North Carolina

89.6

1.4

District of Columbia

124.9

19.4

North Dakota

94.6

-2.8

Florida

90.6

-1.8

Ohio

100.7

0.6

Georgia

90.0

-0.5

Oklahoma

87.3

1.7

Hawaii

113.9

n.a.

Oregon

99.5

1.4

Idaho

89.0

-7.7

Pennsylvania

100.3

5.4

Illinois

107.7

4.5

Rhode Island

101.3

-2.2

Indiana

96.6

0.3

South Carolina

84.9

-3.9

Iowa

102.5

7.2

South Dakota

92.9

-1.0

Kansas

98.0

4.5

Tennessee

89.9

2.5

Kentucky

89.2

-5.7

Texas

87.1

-0.4

Louisiana

86.8

-3.7

Utah

84.8

-14.2

Maine

94.0

2.4

Vermont

94.9

-6.2

Maryland

109.4

-3.4

Virginia

101.2

7.9

Massachusetts

114.0

5.8

West Virginia

89.4

4.8

Michigan

102.2

1.5

Washington

101.5

1.8

Minnesota

104.7

3.8

Wisconsin

101.1

1.4

Mississippi

81.6

-4.8

Wyoming

95.8

-2.5

Missouri

96.8

0.3

-16-

e.g., Mississippi (81.6), and South West, e.g., New Mexico (83.6), where warmer weather and less population density reduces housing costs. The Midwestern and North Central states remain in the middle.

With respect to changes over time, the pattern remains much the same as in 1977. Living costs in Massachusetts, Connecticut, District of Columbia, Michigan, Illinois, and Washington State which were rela- tively high in 1977 now are even higher. And the lower cost of living areas such as Kentucky, Louisiana, New Mexico, and Wyoming now are even lower. Part of this change over time reflects the heavier weight given to rural prices in rural states than in the 1977 study (and vice versa). But part of the change may be related to the change from the earlier oil boom in the southwest to a less vigorous growth in that region as oil prices fell later in the 80's (e.g., Texas -.4, New Mexico -12.1, Arizona -11.3). It is also only more recently with the lower oil prices and industrial recovery from 1985-88 that increases in the cost of living have begun to occur in Massachusetts (+5.8), Virginia (+7.9) and parts of the midwest (Pennsylvania +5.4, Illinois +4.5).

By Counties. Differences in the cost of living among counties in Illinois are shown in Table 2. The regression equation (5) is used to predict these differences based on the per capita income in each county from the 1980 Housing Census, and the change in population from 1980 to 1986 in each county as shown in Appendix D. The index then is normal- ized with a state-wide population weighted mean of 100. The same method could be used in other states.

Table 2 Cost of Living Differences Among Counties in Illinois, 1988

Percent

Percent

Index

Change

Index

Change

County

1988

1977-88

1988

1977-88

Adams

81.7

-12.1

Lee

93.9

1.6

Alexander

80.8

-0.5

Livingston

94.2

2.0

Bond

86.1

-2.3

Logan

93.0

1.4

Boone

94.0

-7.3

McDonough

89.6

-3.4

Brown

84.0

-3.7

McHenry

100.6

-1.1

Butrsu

95.3

4.8

McLean

96.3

-0.9

Calhoun

83.1

-4.1

Macon

97.6

5.8

Carroll

91.1

1.2

Macoupin

86.0

-2.3

Cass

90.1

2.5

Madison

101.3

12.9

Champaign

93.9

-4.0

Marion

86.2

-0.7

Christian

90.3

2.2

Marshall

97.2

6.6

Clark

86.9

-1.3

Mason

95.4

5.1

Clay

81.0

-3.8

Massac

84.3

-2.1

Clinton

86.4

-6.1

Menard

94.3

3.3

Coles

80.8

-12.8

Mercer

91.2

1.2

Cook (Chicago)

102.2

3.8

Monroe

94.2

-2.7

Crawford

80.3

-8.3

Montgomery

84.6

-2.4

Cumberland

84.3

-1.8

Morgan

92.8

-0.3

Dekalb

95.3

-4.4

Moultry

90.1

0.0

Dewitt

95.9

6.8

Ogle

96.3

0.4

Douglas

83.4

-8.2

Peoria

101.6

8.8

DuPage

Ui.o

3.8

Perry

87.5

-1.3

Edgar

88.3

0.1

Piatt

98.0

6.1

Edwards

82.4

-2.3

Pike

84.4

-2.1

Effingham

89.5

-1.7

Pope

77.8

-6.2

Fayette

85.0

-1.5

Pulaski

77.0

-5.2

Ford

92.8

1.2

Putnam

96.7

4.4

Franklin

84.1

-0.6

Randolph

89.9

-0.7

Fulton

92.7

4.4

Richland

88.0

-0.9

Gallatin

82.0

-2.7

Rock Island

97.8

1.5

Greene

84.5

-13.5

St. Clair

109.5

24.7

Grundy

111,0

27.5

Saline

84.2

-1.9

Hamilton

82.7

-0.5

Sangamon

97.6

3.5

Hancock

86.7

-0.4

Schuyler

87.6

-1.1

Hardin

78.4

-4.8

Scott

87.1

-1.3

Herderson

87.2

-2.2

Shelby

87.4

-0.8

Henry

94.8

3.3

Stark

92.4

5.0

Iroquois

90.7

0.2

Stephenson

92.9

-2.0

Jackson

87.0

-7.4

Tazwell

99.0

3.9

Jasper

83.8

-5.2

Union

85.2

-1.3

Jefferson

86.4

-1.2

Vermilion

90.7

2.4

Jersey

88.2

-2.9

Wabash

88.8

3.1

Jo Daviess

90.8

1.2

Warren

91.0

1.0

Johnson

72.1

-18.1

Washington

88.4

-0.5

Kane

98.5

1.1

Wayne

84.7

-1.5

Kankakee

92.8

-0.9

White

87.8

3.3

Kendall

103.5

-3.4

Whiteside

93.6

1.5

Knox

96.4

6.4

Will

96.3

1.2

Lake

111.6

9.4

Williamson

84.4

-4.0

LaSalle Lawrence

89.1

U

Winnebago Woodford

M

"1:1

-18-

These results show a 45 percent variation among counties, from a high of 111.9 in DuPage in the Chicago suburbs and 102.2 in Chicago itself (Cook) to lows of 72.1 in Johnson and 77 in Pulaski and Pope counties. This reflects large urban-rural differences resulting pri- marily from differences in the cost of housing. They are quite com- parable to the 50 percent or so differences in the cost of living among the state averages.

Over time, the cost of living relative to the state wide average has risen in Chicago (Cook) (+3.8%), Chicago Suburbs (e.g., DuPage + 3.8% and Lake +9.4%), and in Peoria (+8.8%). But it has fallen to still lower levels in Johnson (-18.1%), Adams (-12.1%), Coles (-12.8%), and other rural counties adversely affected by the farm recession. The effects from the economic recovery since 1985 and the lower price of the dollar have been felt much more slowly in the farm economy.

V. Conclusions There are large differences of 53 percent in the cost of living among states and of about 45 percent within states. The basic pattern of differences between higher costs in Eastern Seaboard urban and industrial areas and lower costs in Southern and rural areas does tend to persist over time. This is largely because the larger urban areas and bedroom suburbs are typified by higher residential land costs, and higher fuel and other housing costs, and also by higher incomes, a basic pattern that has not changed drastically. There may also be some nonmonetary benefits of living in these areas that at least

-Im- partially justify some of the cost differences. But over time recent changes in the geographical patterns appear to be related to the 1985-88 industrial recovery affecting the northeast, lower oil prices affecting the south in a different way, and the continuing farm reces- sion. In 1980-85 the industrial states were hurt more severely than the oil producing and western states. But prices appear to have been somewhat inflexible downward there, and these areas also recovered more quickly than the agricultural states and rural areas, where land and housing prices remain somewhat lower.

Part of the income differences among areas — roughly a third — are purely nominal differences in monetary salaries, given that there are differences in the cost of living. In the absence of a money illu- sion, employers as well as employees interested in maintaining a parity between services that are purchased or provided in different areas within states or between states must make some kind of adjust- ment implicitly for differences in the cost of living as well as in nonmonetary amenities. A geographical cost of living index is one step toward making such adjustments somewhat more explicit.

-20-

Ref erences

Alonso, William, and M. Fajars (1970), "Cost of Living by Urban Size," Institute of Urban and Regional Development, University of California, Working Paper 128 (Berkeley), 19 pp.

Alonso, William (1970), "The Economics of Urban Size," Regional Science Association Papers, XXVI, pp. 67-83.

Ando, Albert, and Franco Modigliani (1963), "The 'Life Cycle' Hypoth- esis of Saving: Aggregate Implications and Tests," American Economic Review, Vol. LIII (March), as reprinted in AEA Readings in Business Cycles, pp. 398-426.

Barro, Stephen M. (1981), "Educational Price Indices: A State of the Art Review," AIL) Policy Research, 1701 K Street, N.W., Washington, D.C. (March), pp. 1-41.

Haworth, C. T. , and D. W. Rasraussen (1973), "Determinants of Metro- politan Cost of Living Variations," Southern Economic Journal, XL (October), pp. 182-92; J. P. Mattila (1976), "Comment," SEJ, XLII (April); Haworth & Rasraussen (1976), "Reply," ibid.

Hickrod, G. Alan et al. (1987), "Documenting a Disaster: Equity and

Adequacy in Illinois School Finance, 1973 through 1988," MacArthur/ Spencer Series No. 4, Center for the Study of Educational Finance, Illinois State University, Normal, Illinois (December), 43 pp.

Israeli, 0. (1977). "Differentials in Nominal Wages and Prices Between Cities," Urban Studies, Oct. 1977, 14, pp. 275-90.

McMahon, Walter W. , and Carroll Melton (19 78), "Measuring Cost of

Living Variation," Industrial Relations, Vol. 17, No. 3 (October), pp. 324-332.

McMahon, W. W. and Carroll Melton (1977), "A Cost of Living Index for Illinois Counties and School Districts," in Perspectives on Illinois School Finance, Carol E. Hanes, ed., Illinois Office of Education, Springfield (November 1977), pp. 74-113.

Simmons, James et al. (1973), Florida Cost of Living Research Study, Tallahassee, Fla.: State University of Florida.

Sherwood, Mark (1975), "Family Budgets and Geographic Differences in Price Levels," Monthly Labor Review, XCVIIi (April), pp. 8-15.

U.S. Bureau of Labor Statistics (1988), Monthly Labor Review, Various Issues, U.S. Department of Labor, Washington, D.C.

-21-

U.S. Department of Commerce (1980), Census of Population, Bureau of the Census, Washington, D.C.

U.S. Department of Commerce (1980), Census of Housing, Bureau of the Census, Washington, D.C.

U.S. Department of Commerce (1987), Survey of Current Business, Various Issues, Washington, D.C.

U.S. Department of Commerce (1988), "Current Population Reports," Bureau of the Census, Series P-25, No. 1010, Washington, D.C.

Ward, James G. (1987), "The Concept of Adequacy in Illinois School

Finance," MacArthur/Spencer Series No. 5, Center for the Study of Educational Finance, Illinois State University, Normal, Illinois, 21 pp.

Ward, James G. (1988), "City Schools, Rural Schools," MacArthur/Spencer Series No. 6, Center for the Study of Educational Finance, Illinois State University, Normal, Illinois, 8 pp.

D/292

-22-

Footnote

*The author is Professor of Economics, and of Education, at the University of Illinois at Urbana-Charapaign. He is greatly indebted to Wenhui Hu for constructive suggestions and for valuable research assis- tance. He is also indebted to Alan Hickrod and James Ward, as well as to the MacArthur/Spencer Foundation who supported work on this project

D/292

teCKMAN

IINDERY INC.

JUN95

n,i i„ ttmf N MANCHESTER, INDIANA 46962