Gasoline Demand Analysis

 GASOLINE DEMAND ANALYSIS

Applied Economics

MGE 302

David Ennocenti

Dr. Hutton

Gasoline Demand Analysis.max

PROBLEM STATEMENT

Since the early 1970's our economy has been plagued by a

combination of high unemployment and high inflation. A condition

identified by economists as stagflation. During this period our

nation, as well as the rest of the world, has had to deal with the

burden of an energy crisis. Of particular importance is our

nation's dependence on petroleum, more specifically gasoline.

What effect the energy crisis has had on economic growth is

evident. In late 1973, there was the Arab oil embargo, then in

1974, there was a tremendous increase in the price of imported oil.

The great and sudden increase in the price of imported crude oil

had a very substantial effect on both the national and inter-

national economy. It helped to create inflation at home, and

transfered tens of billions of dollars from the oil-consuming

nations to the oil-producing nations.@

In the early 1970's William E. Simon, the nation's energy czar,

wanted to see if more generous reliance on the classic market

mechanism-price as the means of equating demand with supply -

would help solve the energy crisis. His goal was to cut gasoline

consumption by 20 to 30 percent.

In order to see if this goal is feasible it is necessary to

do a demand analysis. A demand analysis serves two major manager-

ial objectivess (1) it provides the insights necessary for the

effective manipulation of demand, and (2) it aids in forecasting

sales and revenues.

How consumption of gasoline will be effected by price increases

depends on it's price elasticity of demand.

Gasoline Demand Analysis.max

Price elasticity of demand is defined as the ratio of the per-

centage change in quantity demanded to the percentage change in

price, assuming all other factors influencing demand remain un-

changed. (i.e., ceteris parabus)

Before making a hypothesis as to the nrice elasticity of

demand for gasoline, I will first take into consideration factors

that govern the size of the coefficient. These factors are (1) the

number and closeness of substitutes for the commodity (2) number

of uses of the commodity and (3) expenditures on the commodity.

The more available and better the substitutes for a commodity,

the greater its price elasticity is likely to be. In the case

of gasoline there are no substitutes to speak of ( with the recent

exception of gasahol and even that is made mostly from gasoline.)

The greater the number of uses of a commodity, the greater

its price elasticity. Gasoline has one use, that is to run engines.

The greater the percentage of income spent on a commodity,

the greater its elasticity is likely to be. For example, the de-

mand for automobiles is likely to be much more price elastic than

that for gasoline.

Given these factors c!„..31- 416 that the

price elasticity of demand for gasoline is inelastic.

Demand studies of gasoline have made the following con-

clusions. Alan Greenspan of Townsend-Greenspan and Co., be-

lieves that the elasticity of gasoline demand is about -0.4.

A preliminary estimate from the Transportation Departments re-

search center shows elasticity in the -0.2 range. A model de-

veloped by Professor Hendrik S. Houthakker of Harvard and

Philip K. Verleger, Jr. of Data Resources, Inc., puts short-run

elasticity at -0.3.

Since demand is hypothesized to be price inelastic, if price

were to increase so would total revenue increase.

a

Gasoline Demand Analysis.max

Empirical evidence supports the hypothesis that demand for

gasoline is price inelastic. The price per gallon of regular grade

gasoline, excluding taxes, increased in the second quarter of 1980

by more than 40 percent over the price of the same quarter in 1979.

Figures published in Business Week's Corporate Scoreboard shows

that in the middle of one of the sharpest recessions on record,

oil companies increased their net aftertax earnings in the second

quarter 1980 by some 32% over the year ago figure while the rest

of the U.S. industry was suffering a profit decline of 18%. 0

The coefficient on income elasticity of demand measures the

percentage change in the amount of a commodity purchased per unit

of time resulting from a given percentage change in a consumer's

income, ceteris paribus.

When the coefficient is negative, the good is inferior. If

elasticity is positive, the good is normal. A normal good is

usually a luxury if its income elasticity of demand is greater

than one, otherwise it is a necessity. Therefore, gasoline's

044, t•

income elasticity should be since it is neither an in-

ferior nor luxury good.

Gasoline Demand Analysis.max

MODEL DEVELOPMENT

In developing a demand function price and disposable income

are usually included as factors influencing consumption of a com-

modity. Since gasoline is a normal good. as the price of gasoline

increases, the amount of gasoline demanded should decrease. As

per capita disposable income increases the quantity of gasoline

demanded could be expected to increase. Again this is because gas-

oline is a normal good.

In addition to price and disposable income, there are other

factors which effect the consumption of gasoline. The number of

registered motor vehicles is included in this demand analysis. As

the number of registered motor vehicles increases (decreases) the

quantity of gasoline demanded should increase (decrease).

ImitaAnuch as there are no substitutes for gasoline one can-

not be included in this model. However, although people cannot

find another commodity in which to run their cars, people can use

other means of transportation. To take account of this factor,

urban transit systems revenue is included in the demand analysis.

As this increases the demand for gasoline should decrease.

At the start of the third quarter 1974, President Richard Nixon

initiated a mandatory fifty-five mile per hour speed limit. This

was introduced as an effort to improve the fuel efficiency of

automobiles and curb demand for gasoline. This variable is entered

in the demand function as a dummy variable, taking a value of one

for years in which the fifty-five mile per hour speed limit was used

and a value of zero for all others.

A trend variable is included in the demand function to ex-

plain the change in quantity demanded over time. As this value

increases demand should increase.

Gasoline Demand Analysis.max

Since the data in this demand analysis is quarterly, seasonal

trends are a counted for by the use of dummy variables. The de-

mand for gasoline is typically lowest in the first quarter, higher

in the second, reaches its peak in the third quarter, and drops

to a level in the fourth quarter which is slightly lower than the

second quarter and higher than ther level of the first quarter.

In summary, the factors considered for this analysis are

price, per capita disposable income, number of registered motor

vehicles, urban transit revenue, the introduction of the fifty-

five mile per hour speed limit, a trend variable, and dummy var-

iables to measure seasonal effects. All of these variables were

used in the statistical analysis.

The demand function to be tested is:

Y1= a + bXi + bXt + bXp + bXu + bXm bXr + bXa + bXb + bXc + e

Where: Y1 = Quantity of gasoline demanded in millions of barrels.

Xi =Per carita disposable income in 1967 dollars.

Xt = Quarterly trend variable form the first quarter 1970

to the fourth quarter 1978.

Xp = Price per gallon of regular grade gasoline in 1967.

Xu = Urban transit revenue in 1967 dollars.

Xm - Dummy variable for the introduction of the fifty-

five mile per hour speed limit.

Xr = Registered motor vehicles in millions.

Xa = Seasonal trend variable (1) for the first quarter of

each year.

Xb = Seasonal trend variable (1) for the second quarter

of each year.

Xc = Seasonal trend variable (1) for the third quarter of

each year.

e = The disturbance term.

Gasoline Demand Analysis.max

(Data sources for the nrevious cage are listed in the appendix)

Given the previous information the exrected signs of the co-

efficients are as follows:

It = a + bXi bXt bXp - bXu - bXm + bXr - bXa + bXb + bXc.

The reason for the expected negative sign for Xa is, this

is the quarter in which gasoline consumption reaches its lowest

point. Xb should be positive due to the upswing in demand in the

second quarter over the first quarter. In the third quarter, gas-

oline consumption reaches its peak. The ore, Xc should be

positive.

6

9

Gasoline Demand Analysis.max

EVALUATION OF RESULTS

The method of analysis used in this demand study is the

least squares regression method. The multiple regression equation

identifies the best fitting line based on the method of least

squares.

The multiple regression model used to estimate the demand

function for gasoline yielded the following results:

Y1 = 146.372 + .017Xi ¨ .018Xt - 372.744Xp

(.710) (.140) (-7.515)

-.014Xu + 2.746Xm + 4.058Xr - 54.729Xa

(-3.186) (.539) (3.713) (-11.484)

+ 6.920Xb 21.540Xc

(1.970) (7.635)

Where, the values in parentheses represent t-ratios.

The only inconsistency between the hypothesized results and

the actual results is the positive value of the beta coefficient

for Xm, the variable for the fifty-five mile per hour speed

limit. This may have been caused by problems I will discuss

later in the case.

The coefficient of determination measures the proportion of

the variation in the independent variable (quantity of gasoline

demanded) that is explained by the regression line (the independent

variables) This is calculated by dividing the explained sum of

squares,(regression sum of squares) by the total sum of squares.

This value is calculated to be .99407 and is designated by R square.

In other words, the regression equation explains 99.4 percent of

the variation in the quantity of gasoline demanded.

A value of .99703 is calulated for r, the coefficient of

correlation. The sample coefficient of correlation, r, is some-

what biased as an estimator, with an absolute value which is too large.

Gasoline Demand Analysis.max

An unbiased estimator for the coefficient of determination,

R square, is adjusted R square.

e This value is .99201.

Another method of evaluating the explanatory power of the

regression equation is the use of the F-test. This is used to

test the null hypothesis that all the regression coefficients are

zero.

Critical F at the 5 percent level of significance for 9/26

degrees of feedom is 2.27. The F-ratio is the mean square re-

gression divided by the mean square residual. If the calculated

F is greater than the critical F, the decision is to reject the

null hypothesis that there is no significant relationship be-

tween the dependent variable and the independent variables of the

regression equation.

The calculated value of F is 483.9. Since this value is

greater than the critical value of 2.27, we reject the null

hypothesis that there is no significant relationship between the

dependent variable and the independent variables.

Testing the individual beta's of the coefficients is ac-

complished by the use of the t-ratio. This statistical test is

used to determine whether the smple value b for each of the coef-

ficients is significantly different from zero. The null hypothesis

is that B=0 and the alternate hypothesis is that BX0. The sample

statistic will have a t distribution with (n-k-1) degrees of

freedom. Where nz the number of observations and k = the number

of independent variables. Degrees of freedom for this analysis

is (36-9-1) 26. Critical values for t at the 5 percent level

of GiunifinannP nnd 2 cipgrppR of frepcinm ArP 2J)56

and +2.056.

If the calculated value of t is either less than -2.056 or greater

than +2.056, the decision is to reject the null hypothesis and

conclude that a statistically significant relationship exists be-

tween the quantity of gasoline demanded and the independent varihip.

Gasoline Demand Analysis.max

The test statistic is b-0 divided by the standard error of beta.

The tests for the individual beta's are as follows.

For per capita disposable income t = .710 which means b is not

significantly different from zero.

The quarterly trend variable representing time, Xt, has a

t-ratio of .138 and this is not statistically significant.

For the price of gasoline t = -7.515. This is statistically

significant.

Urban transit revenue has a t-ratio of -3.186. This is

significantly different form zero.

For the introduction of the fifty-five mile per hour speed

limit, the dummy variable Xm, b has a t-ratio equal to .539. This

is not statistically significant.

Registered motor vehicles has a t-ratio of 3.713. This is

significantly different from zero.

Seasonal trend variable Xa has a t value of -11.484. This

is significantly different from zero.

Seasonal trend variable Xb has a t-ratio equal to 1.970.

This is not statistically significant.

Seasonal trend variable Xc has a t of 7.635 and is statistic-

ally significant.

In summary, significant variables in the equation are, Price

per gallon, urban transit revenue, registered motor vehicles,

seasonal trend variable Xa, and seasonal trend variable Xc.

Variables not statistically significant are, per capita disposable

income, time, the fifty-five mile per hour speed limit, and

seasonal trend variable Xb. These variables may have tested to be

insignificant due to problems that can occur with multiple

regression analysis which I will define.

9

Gasoline Demand Analysis.max

10

"Otlix , Cry0

There are four assumptions of multiple c-errrtrnti7511 analysis

these are that : (1) all variables involved in the analysis are

random variables (2) the relationships are all linear (3) the

conditional variances are all equal and (4) the conditional

disturbances are all normal. These requirements are quite stringent

and are seldom completely satisfied in real data situations.

However, multiple correlation analysis is quite robust in the sense

that some of these assumptions, and particularly assumption number

four, can be violated without serious consequences in terms of the

validity of the results.

Problems which can violate some of these assumptions are,

autocorrelation, heteroscedasticity, specification and measurement

errors, and multicollinearity.

Autocorrelation refers to the existence of a significant

pattern in the successive values of the error term. Positive

autocorrelation is inferred whenever successive positive (negative)

disturbances tend to be followed by disturbances of the s2me sign.

Negative autocorrelation is inferred whenever successive positive

(negative) disturbances tend to be followed by disturbances of the

opposite sign. The overall growth of the economy coupled with

business cycles causes most economic time series to have an over-

all upward trend with periodic upturns and downturns around this

trend producing positive autocorrelation. qp Seasonal trend

variables, which are included in this analysis, can help eliminate

this problem.

The Durbin-Watson test is reported in the comnuter printout

and is a test for autocorrelation. A Durbin-Watson value of less

than 1.5 is usually a indication that positive autocorrelation,is

present. The value reported in this regression is 2.482, which

clearly indicates there is no positive autocorrelation nroblem.

Gasoline Demand Analysis.max

Critical values for Durbin-Watson are 1.5 and 2.5. Values outside

these extremes usually imply autocorrelation, The reported value

of 2.482 is close to 2.5. This high value may have been caused

by the number of runs of signs (25) of the residuals. Any further

runs of signs may have caused autocorrelation problems. Invest-

igation of the residuals shows areas where the disturbances are

followed by disturbances of the same sign, as is the case with

positive autocorrelation. For this reason, and the fact that

Durbin-Watson is below 2.5, I feel that autocorrelation is not a

serious problem with this regression.

Homoscedasticity is the assumntion that the error terms,

which are independent random variables, have a uniform variability

about the regression line. The lack of a uniform variance of the

error terms is known as heteroscedasticity. Observation of the re-

siduals gives no indication of heteroscedasticity. The residuals

appear to be randomly distributed about the regression line.

Measurement errors are possibly present in this regression

equation. These errors occur in the collection process of the

data used. For example, the nrice per gallon of gasoline was

made from a random sample of fifty-five U,S. cities at midmonth

prices. This sample may be a different value than the true price.

A problem that is present in this model is multicollinearity.

Multicollinearity is indicated whenever two or more expanatory

variables are highly correlated and the standard errors of their

regression coefficients become large, causing the t-ratios to go

down. (D Multicollinearity can also cause sign changes of the

coefficients.

Mentioned earlier was the fact that the coefficient for the

fifty-five mile per hour speed limit should have been negative but

the value in the regression became positive.

Gasoline Demand Analysis.max

12

The explanation for this occurrencemulticollinearity, This

variable is highly correlated with time, price, and registered motor

vehicles. Their correlation coefficients with the fifty-five mile

per hour speed limit are .866, .861, and.839, respectively. The

relationship between these variables may have caused the stand-

ard error to increase causing the t-ratio to prove the speed limit

variable is not significant. Multicollinearity may also have

caused the sign change with this variable.

Other variables which proved to be insignificant were , per

capita disposable income, time, and seasonal trend variable Xb.

The trend variable representing time is highly correlated

with income, price, registered motor vehicles, and the variable

for the speed limit. Their respective correlation coefficients are

.912, .816,.991, and .866. This clearly indicates that the trend

variable for time is highly correlated with these other variables.

The beta value for time is .180 and its standard error is 1.294.

The standard error may have increased due to multicollinearity

causing the t-test to be an unreliable indicator of the statistical

significance of the trend variable for time.

Correlation coefficients between income and the variables

time and registered motor vehicles are .912 and .921, respectively.

As with the trend variable for time, the beta value for income may

have proven insignificant due to multicollinearity. Beta for income

is .017 and its standard error is .024. The error term may have beet

inflated thus, causing the t-ratio to decrease and prove the

variable to be insignificant.

Seasonal trend variable Xb, representing the second quarter

has a beta which is not significantly different from zero. Unlike

the three pervious variables, multicollinearity had no effect on

the t-ratio of the trend vari,qhlp Yh.

Gasoline Demand Analysis.max

This is evident from the fact that its highest correlation

coefficient with any variable is -.333 this is with both the

trend variables Xa and Xc. Deductive reasoning can lead to the

conclusion that the beta for trend B is in fact not statistically

significant. This is because the demand for gasoline in the first

quarter is at its lowest annual point. In the third quarter

demand reaches its peak. Since the second quarter, as well as the

k beta

fourth, falls between these two extremes4can be hypothesized to

be statistically insignificant. This is because the second and

fourth quarters would approximate the average of the four quarters

more closely than the first and third quarters (the two extremes)

Ommission of this variable would result in multicollinearity

problems between the trend variables A and C.

In summary, multicollinearity may have affected the following

variables time, income, and the fifty-five mile per hour speed limit.

However, the presence of high intercorrelation among the independent

variables does not necessarily invalidate the use of the regres-

sion equation for prediction purposes. Provided that the inter-

correlation pattern among the explanatory variables persists into

the future, the equation can produce reliable forecasts of the

value of the dependent variable. 0

It can reasonably be expected that the fifty-five mile per

hour speed limit will continue into the future as well as an in-

crease in the number of motor vehicles. The trend variable, time,

will obviously also increase into the future. Therefore, mul-

ticollinearity should not invalidate the use of this regression

equation for prediction purposes.

95 percent confidence intervals for the individual beta

coefficients are given in the computer output. The papameter

beta can be estimated by constructing the following confidence in-

terval.

13

Gasoline Demand Analysis.max

b (+) or (-) t(standard error of beta)

Degrees of freedom associated with t are (n-k-1). Where,

k is the number of independent variables and n is the number of

observations. With 26 degrees of freedom the value for t is 2.056.

Confidence intervals are as follows:

Income Xi .032 to .065

Time Xt 2.481 to 2.840

Price Xp - 475.0 to-271.0

Urban Transit Revenue Xu - .022 to ,005

Speed Limit Xm - 7.73 to 13.22

Registered Motor Vehicles Xr + 1.8. to 6.30

Trend A Xa - 64.5 to-44.9

Trend B Xb - .301 to 14.141

Trend C Xc 15.7 to 27.3

The standard error of estimate can be used to establish

prediction interval for the dependent variable given specific

values of the independent variables. This value has been cal-

culated by the computer to be 4.844.. A prediction interval can

be estimated by the normal distribution or by the t distribution.

When using the normal distribution a 95 % confidence interval

would be, the Y estimate minus (2x4.844) to the Y estimate plus

(2x4.844). When using t, the 95% confidence interval with 26

degrees of freedom is, the Y estimate minus (2.056 x 4.844) to

the Y estimate plus (2.056 x 4.844).

As expected, income elasticity of demand is estimated in

this analysis to be inelastic. The value of .086 mens that as

income increases (decreases) by 1% demand will increase (decrease)

by .086%. This value, though quite low, is reasonable for gas-

oline. Since gasoline is not an inferior good the elasticity should

be positive. Luxury goods have an elasticity which is greater than

nn e.

Gasoline Demand Analysis.max

Gasoline, which most people view as a necessity, should have an

elasticity greater than zero and less than one.

The validity of this calculated value is questionable due

the

to the fact that/theta coefficient did not prove to be significantly

different from zero. As mentioned earlier, this may have been

caused by multicollinearity.

The price elasticity of demand in this analysis is estimated

to be .15. Since this value is less than one, price elasticity of

demand is inelastic. This is consistent with what it was

hypothesized to be. The value of .15 is lower than other estimates

of .4 and .3 but is close to the .2 estimate made by the Transport-

ation Department's research center. The elasticity measure is

statistically significant because the beta coefficient for price

tested to be significant.

In order to use price as a mechanism for cutting demand by 30

percent, price would have to be increased by 200 percent or. in

other words, nrice wouldhave to be tripled. To cut demand by 20

percent would call for an increase in price of 193 percent. Which

means price would have to be more than doubled.

Controlling demand with such substantial price increases may

come at the exrense of serious consequences. Tremendous Price

increases could bring about a massive distribution of profits,

as in 1974, from oil consumers to the oil producers. This massive

shift in profits could have serious implications for the United States

economy. It could:

(1) Lower the rate of economic growth as more and more

companies find it increasingly difficult to acquire funds for

capital investment either form internally generated sources or

from financial markets.

(2) Lead to continued loss of economic efficiency within the

corporTte sector as oil company managements, with more money than

IS

Gasoline Demand Analysis.max

they know what to do with, take on unfamiliar projects, while

other companies have to abandon some ventures because of lack of

funds.

(3) Exacerbate inflation as companies seek to boost their de-

clining profit margins by raising prices. 0

These assumntions lead to the conclusion that relying on price

as a means of reducing gasoline consumption could create further

economic problems. Perhaps the solution to our energy crisis

lies not in price increases alone, but in a combination of al-

ternatives. Some of which I will now consider.

The number of registered motor vehicles has increased by

37.5% from 1970 to 1978. The results of this regression equation

indicates that this factor has had the effect of incrensing the

demand for gasoline. Since the mid-seventies automobile manufact-

urers have been improving the fuel efficiency of automobiles.

Further improvements seem to be called for in order to help cut

gasoline consumption.

The beta coefficient for the variable representing the fifty-

five mile per hour speed limit indicates that this has not helped

decrease demand. As mentioned earlier, this value may have taken

on a positive value and proved to be insignificant due to multi-

collinearity nroblems. In any case, additional speed reductions

may be necessary in order for this variable to have any significant

effect in reducing demand. aeot

IrmA0Le.

lao 4-464,4-

According to this analysis urban transit systems has hd a

decreasing effect on gasoline demand. Increased use of public

transportation should be encouraged in order to decrease the demand

for gasoline.

The fact that gasoline has no substitutes is one of the factors

which causes its price elasticity to be inelastic.

16

Gasoline Demand Analysis.max

The development of an alternate source of energy seems to be an

imperative solution to the energy crisis. In July of 1980,

Fresident Jimmy Carter signed into law a synthetic fuels program.

This program is supposed to spur the production of 500,000 barrels

of "synfuels" a day by 1987 and two million a day by 1992. Until

the time this production takes place, measures will have to be taken

to curb the demand for gasoline. Due to economic difficulties

which may arise, reliance on substantial price increases does not

seem feasible.

In summary, moderate urice increases together with further

speed reductions, improved fuel efficiencies of new automobiles,

increased use of public transportation, and efforts to develop

an alternate source of energy may be one possible soulution to our

current energy crisis.

IT

Gasoline Demand Analysis.max

CONCLUSION

Variables that could be added to the equation are, total number

of miles driven by all vehicles, the number of licensed drivers,

and the fuel efficiency ratings of motor vehicles.

The number of miles driven by all vehicles was not found in

quarterly data in the sources used for this analysis. The beta

coefficient for this variable could be expected to be positive,

seasonal, and highly correlated with demand. The number of miles

driven per quarter could be used instead of the trend variable

representing time.

As the number of licensed drivers increases demand for gasoline

could be expected to increase. The beta coefficient for this

variable would be hypothesized to be positive.

Fuel efficiency ratings of automobiles has been a recent

issue of concern. Due to this, data may not be available for all

the years used in this analysis. One method that could be used

to include this variable in the model would be to divide the

number of miles driven by the number of barrels of gasoline de-

manded, then divide this figure by the number of motor vehicles.

This figure would be the average miles per barrel of gasoline per

motor vehicle. As this figure increases demand could be expected

to decrease. Therefore, the beta coefficient would be hypothisized.

to be negative.

Advertising expenditures were not included in this model al-

though they very well could have been. However, I do not feel that

advertising has had much effect in influencing demand consumption

in the 1970's. In marketing terminology, gasoline is in the

maturity stage of the product life cycle.

Gasoline Demand Analysis.max

In this stage marketing expenditures are directed toward brand

preference and services.

3

This would effect individual companies

but not the industry as a whole.

An advertising variable could have been included which may

have influenced gasoline demand. The government has been active

in the demarketing of gasoline consumption throughout the 1970's.

It would be interesting to see if this has had any effect on de-

creasing gasoline consumption. This variable would be hypothesized

to have a negative beta coefficient.

I feel this model would be useful in predicting gasoline

demand within a 95% confidence interval. The reasons for this are

the high R square value of .994 and the overall F-test which

proved to be significant. However, as with all regression models,

limitations do exist with the use of this model for a means of

predicting.

Common Limitations are:

(1) A value of the dependent variable cannot be legitimately

estimated if the values of the independent variables are outside

the range of values which served as the basis for the regression

equation.

(2) If the estimate of the dependent variable involves the

prediction of a result which has not yet occured, the historical

data which served as a basis of the regression equation may not

be relevant for future events.

(3) The use of a prediction interval is based on the as-

sumption that the conditional distributions of the dependent

variable are normal and have equal variances. k.9

• AA

11116 1d25 L 1UrrIltraUlOn iS concerneu wiuu autocorreiaLun anu

heteroscedasticity. As mentioned earlier I don't feel these problems

are present in this model.

14

Gasoline Demand Analysis.max

Most of the variables in this regression should not be

subject to the first two problems. However, the price of gas-

oline could present a nroblem if price should suddenly and sub-

stantially increase.

For example; Alan Greenspan, whose model WPS mentioned in

the introduction, feels his estimates are valid only up to 15

cents above the current price. His model was made in 1973 when

the annual average price per gallon was 27.5 cents.

Within the limitations of any regression equation, I feel

this model would be useful in predicting the demand for gasoline

within a 95% confidence interval.

20

Gasoline Demand Analysis.max

Bibliography

1) Principles of Microeconomics ; Edwin Mansfield ; Second

Edition ; Page 394

2) Principles of Microeconomics ; Readings, Issues, and Cases ;

Edwin Mansfield ; Second Edition ; Pg. 49

3) Managerial Economics ; James R. McGuigan and R. Charles Moyer ;

Second Edition ; Pg 100

4) Principles of Microeconomics ; Readings, Issues, and Cases ;

Edwin Mansfield ; Second Edition ; Pg 49

5) Business Week ; August 18, 1980 ; Pg 84

6) Business Statistics ; Leonard J. Kazmier ; First Edition Pg 304

7) Business Statistics ; Leonard J. Kazmier ; First Edition Pg 916

8) Managerial Economics ; James R. McGuigan and R. Charles Moyer;

Second Edition; Pg 84

9) Managerial Economics ; James R. McGuigan and R. Charles Moyer;

Second Edition ; Pg 87

10) Managerial Economics ; James R. McGuigan and R. Charles Moyer;

Second Edtion ; Pg 87

11) Business Week ; August 18, 1980 ; pg 84

12) Wall Street Journal ; July 11, 1980 ; pg 17

13) Principles of Marketing ; Philip Kotler ; First Edition ; Pg 354

14) Business Statistics ; Leonard J. Kazmier ; First Edtion ; Pg 304

15) Principles of Microeconomics ; Readings, Issues, and Cases ;

Edwin Mansfield ; Second Edition ; Pg 50

Gasoline Demand Analysis.max

Appendix

Data Sources

United States Bureau of the Census: Statistical Abstracts

100th Edition, Sept. 1979.

1) Motor vehicle registrations in millions includes, total

automobiles, trucks, and buses.

Survey of Current Business, Volumes 51 through 59, Volume 60

number 8.

1) Disposable personal income in billions of dollars.

2) Total United States population in millions.

3) Price per gallon of gasoline, regular grade, excluding

taxes. Taken from 55 U. S. cities at mid-month.

4) Urban Transit Systems total revenue, from passengers

carried, in millions.

5) Domestic demand of gasoline, in millions of barrels.

6) Consumer Price Index for all items, base year 1967.

(Not seasonally adjusted)

7) Consumer Price Index for Public Transportation, base year

1967 (Not seasonally adjusted)

Gasoline Demand Analysis.max