In That Paper (Intro) They Note That There Are a Lot of Estimates of the Price Elasticity

Total Page:16

File Type:pdf, Size:1020Kb

In That Paper (Intro) They Note That There Are a Lot of Estimates of the Price Elasticity

Senior Project

Department of Economics

“Elasticities of demand for gasoline in the United States”

Presented by: Milana Borisev

December 12, 2007.

Abstract

Motor gasoline is the most notable and desirable product made from refining oil. This research paper estimates the effect of the price change of gasoline on demand of gasoline within the context of a fully specified model of consumer demand for this product. I will also estimate price and income elasticity of demand for gasoline in the US. The data set that I will use in this paper is a panel data set of 50 states and it will consider years 2001-2005. The principal hypothesis that I am going to investigate is how consumers demand for gasoline responds to price and income changes in given period of time. Table of Contents

PART 1: Introduction …………………………………………. Page 2

 Statement of the problem  Literature Review

PART 2: Methodology …………………………………………Page 8

 Data Description  Presentation of the model  Explanation of the model

PART 3: Data sources and description ………………………... Page 11

 Discussion of results  Results summary table

PART 4: Conclusion ……………………………………………Page 13

 Conclusions  Policy Implications  Limitations

References……………………………………………………… Page 16

2 1. Statement of the Problem

Understanding the sensitivity of gasoline demand relative to income and price changes is very important. With U.S. gasoline consumption accounting for about 11 percent of world oil production, the US has been hit hard by its dependence on oil. This dependence is a threat to its national security and to the economy because most of the oil is imported from unfriendly and unstable regions (NRDC 2004.) The growing demand for gas and reduction in domestic production forces the US to import more and more oil each year.

Of the $54 billion trade deficit reported in August 2004, more than a fifth or $12 billion is from imported crude oil.1. From year to year US gas production is decreasing while demand has increased steadily over the 1975-2005 period. The root cause of high gasoline prices is soaring demand, caused in large part by increasingly fuel-inefficient cars and trucks.

The big question is how to decrease dependence on the imported oil and what should be done to make sure that oil is supplied only from the reserves located in the US. I do believe that it is very important for the US because most of the countries that the US imports oil from are economically and politically unstable countries and the risks with doing trade with this countries are too risky. One of the options that has been proposed and that could help in reducing US dependence on crude oil is tax policy. There is evidence that raising the taxes on motor fuel would actually decrease US dependence on crude oil imports and help the nation as a whole2.For example, Alan M.Schneider showed that a tax

1NRDC (October 2004) 2 NRDC (October 2004) 3 of 5 cents per gallon per year over the period 1979-1989 can be expected to produce significant reductions in gas consumption.3

So who would actually pay for the taxes, consumer or supplier? In the short run consumers will pay a higher percentage of the tax because they would not respond to price changes, so demand would be more inelastic than supply. On the other hand, in the long run demand would be more elastic than supply, so suppliers would pay a higher share.

Many studies focus on estimating U.S. consumer’s behavior and responses to changes in gasoline prices. However, much of this research focuses on the 1970s and

1980s. At that time the focus was determining short and long run price and income elasticities. Some authors also added other factors such as inflation; unemployment and interest rate as influences on gasoline demand. Since that time, U.S. gasoline market had changed in many ways, and so may consumer responsiveness to price change of gasoline.

There are number of reasons to believe that the previous period’s price elasticity differs from the current period. Some of the factors that caused this are different behavioral and structural factors, technological improvements, different time periods, wide spread of globalization, growth of multiple income households and per capita income, as well as dependence on automobile usage.4

For example, Polzin and Chu (2005) found that over the past thirty years the share of transit passengers miles traveled relative to other modes has drastically decreased.5 The possible theory behind this finding is that even though gas prices increased over period of time, more efficient vehicles were made and those new vehicles allowed drivers to spend less on gas per mile traveled. In another study, Hughes, Knittel, and Sperling (2006), find that in order to achieve the equivalent reduction in gasoline consumption in 1970s and

3 Schneider (1978)

4 Hughes, Knittel, and Sperling (October 2006) 5 Polzin and Chu. (2005) 4 2000s, gasoline taxes in absolute terms would need to be significantly larger6. This implies that today the demand for gas is more price inelastic in comparison to previous periods, and that, in order to have the same reduction in consumption, prices of gas would have to be much higher today than before.

In this paper I will estimate the effect of the price change of gasoline on demand of gasoline within the context of a fully specified model of consumer demand for this product.

The data set that I will use in this paper is a panel data set of 50 states and it will consider years 2001-2005. The principal hypothesis that I am going to investigate is how consumers demand for gasoline responds to price and income changes in given period of time. Also, I would like to learn how factors, other than prices and income, affect consumption and in what way.

Literature Review

Gasoline demand and its sensitivity to changes in income and prices plays important role for policies that are related to optimal taxation and national security. As noted above there are number of reasons to believe that in the past few periods demand elasticties have changed. For example Hughes, Knittel, Sperling (2006) argued that factors such as technology and transportation played crucial role on US consumer’s behavior and its responsiveness to changes in gasoline prices. They estimate the short run income and price elasticities of gasoline demand using a US aggregate monthly data for the period from

1974 to 2006.Their estimates, presented in Figure 1, show that price elasticity of demand for gasoline for the period 1975-1980 range between -0.21 and – 0.34, which is relatively price inelastic. (Minus sign for elasticity only means that price and consumption are negatively related). Price elasticity of demand for the period from 2001 to 2006, are lower,

6 Hughes, Knittel, and Sperling (October 2006) 5 from -0.034 to -0.077. The conclusion is that the short run price elasticity of demand for gasoline is substantially more inelastic today then it was in previous periods.

The empirical models just explained were estimated for each period using ordinary least squares (Hughes, Knittel, and Sperling 2006).Variables that they used in their model are monthly gasoline consumption, real gasoline prices as city average prices for unleaded regular fuel and personal disposable income. The results of Hughes, Knittel and Sperling’s ordinary least squares experienced problem in estimating price elasticity of demand for gasoline. Price and quantity are jointly determined through shifts in both supply and demand resulting in biased and inconsistent parameter estimates using OLS techniques.7

This problem is very important when trying to compare elasticity estimates between two periods.

Another hypothesis of their study is that as income has grown, the budget share represented by gasoline consumption has decreased making consumers less sensitive to price increases. ( Hughes, Knittel, Sperling 2006). They proposed that if increasing income leads to decrease in the consumers response to gasoline price changes, one would expect the coefficient on the interaction between price and income to have a positive sign.

However, their research lead them to the opposite; the coefficient on the interaction term has negative sign, meaning that on average gasoline consumption is more sensitive to price changes as income rises. I think that one of the reasons why the interaction term has negative sign is that if income level increases so does the number of vehicles per household. When number of vehicles exceeds number of drivers, there is a possibility for drivers in that family to shift to more fuel efficient cars as gasoline prices rise.

Hughes, Knittel, and Sperling (2006) observed changes in the periods from 1975 to

1980 and 2001 to 2006 that are evidence of a structural change in the US market for

7 Hughes, Knittel, and Sperling (October 2006) 6 transportation fuel, shifts in land-use, social or vehicle characteristics and also improvements in technology.

Figure 1 Monthly price elasticity estimates from May 1978 through September 2003.8

Small and Van Dender (2007) estimate the rebound effect for motor vehicles. The main idea of paper is that improved energy efficiency, which leads to cheaper process, makes people increase its use. In this case, when vehicles are made more fuel efficient, it costs less to drive a mile, so vehicles miles traveled increases, causes more fuel to be used than would be the case if vehicles- mile traveled were consistent.9 They concluded that rebound effect diminishes with income, and probably increases with the fuel cost of driving.

A paper written by Dahl (1982) investigated the issue of whether gasoline demand elasticities are constant over different levels of incomes and price. She wanted to see

8 Hughes, Knittel, and Sperling (October 2006) 9 Small and Van Dender( 2007) 7 whether gasoline demand elasticities are same for income and price increases as they are for income and price decreases. Dahl believed that consumers are the one that decide how

much gas their vehicles are consuming by making a decision on what vehicle they want to drive and the intensity of their use. The results that she came up with are that the short run elasticity estimate for price is -.20 and for income.11. Long run elasticity estimate for price is -.98 and for income .50. An additional consistency of elasticity over large price changes that were found in her results indicates that predictions are valid even for large market disruptions.10

A paper written by Nicol (2003) estimated elasticity of demand for gasoline for various households groups.11 Nicol believed that gender of head of households would influence elasticity of demand for gas. Some other independent variables that Nicole included in estimating the model are age of head, consumption of tobacco and a vehicle ownership variable. He concluded that gasoline demand is price inelastic and income inelastic (2003). Family size and housing tenure status also have larger impacts on differences in elasticities across households.

The dynamics and composition of household adjustment to changes in the real price of gasoline using a panel data set of US households is well explained in the model estimated by Puller and Greening (1999). For many years, policy makers and researchers were trying to understand consumer response relative to changes in price of gasoline so they could design effective environmental and energy policy. This paper uses nine years of

US household –level panel data to estimate long-run price elasticites of non-business gasoline demand and also to analyze the nature and timing of household adjustments to changes in gasoline prices.12 Puller and Greening estimate the demand for gasoline in a

10 Dahl (August 1982) 11 Nicol (2001)

12 Steven L. Puller, Lorna A. Greening (1999) 8 one-equation model, and then divide the demand for gasoline into the demand for vehicle miles traveled and the demand for household miles per gallon. The results of the model are consistent with the claim that gasoline demand is relatively price inelastic in the year following a price change. (Puller, Greening 1999).

From the previous studies we can conclude that demand for gasoline in the US is price inelastic and income inelastic. Also, demand is more price inelastic today then it was in previous periods. Other factors, such as, technological changes, number of vehicles per household, production of more fuel efficient cars and availability of different modes of transportation, affect consumption of gas.

In this paper I will estimate price and income elasticity of demand for gas based on a state data. I will also use some other independent variables that will help me better explain changes in per capita consumption of gas. As I already mention, gas taxes are very important factor in determining gas consumption. I will explain effect of tax change on consumption of gas.

2. Formulation of the Model

A demand function can be described as how much of good will be purchased at alternative price of the good and related goods, alternative income levels and alternative values of other variables affecting demand. I will estimate to see what factors effect the consumption of gas and in which direction. The data set consists of 50 US states, it will include period from 2001 to 2005.

From consumer demand theory for normal goods, we know that higher prices of a good, other things constant, force consumers to demand less. With an increase in income level consumers’ purchasing power would increase and they would buy more of that good.

9 If state size increases, meaning more people living in that state, so more gas will be demanded. Urban population is another independent variable that I will use. With more people living in cities, there is an option to use public transportation instead of a car. Also, assuming that the largest percentage of those people that live in the city also work in the same city so they could even walk. As an alternative means of transportation they use bus, trolleybus, train or maybe walk rather than drive, they wouldn’t consume that much gas as they would if they lived outside the city. The last variable that I use in my model is number of cars per person. Some of the families in the US have one car per person; on the other hand some households have one car for the whole family. If, there is a one car per person in the family that live under same roof, that would mean that all of the household members can drive whenever and wherever they want. So each of the members would have to put gas in his or her car, on the other hand if there is only one car in the family that means that only one car would need a gas. Based on my previous explanation I would expect number of cars per person to have positive influence on consumption variable. My model is different from the models used in previous studies about demand for motor vehicles.

This paper estimates the United States demand function for motor gasoline in the

United States, which will take a form similar to the equation below. Two most important factors that effect consumer demand for any good are price of good and their income level. a) I first estimate the model using ordinary least squares:

Consumption = β0 + β1 price + β2 rgdp+ β3 state size + β4 urban population+ β5 number of cars+ + ei b) There is a possibility of problems with the OLS because two or more dependent variables are determined at the same time. In this model, price and consumption are both dependant variables because they are both determined by other variables in the supply and demand equation. In this case model will need more than one equation. c) The OLS can not be used to estimate a simultaneous-equation system. The OLS estimates each equation separate and it does not take into account that equations are

10 part of large system, so the estimates will be biased. Possible way out to deal with simultaneous bias is two stage least squares:

^

Consumption= β 0 + β 1price + β 2 rgdp + β 3 state size + β 4 urban population + β 5 number of cars+ ei

d) An ideal instrumental variable for determining gasoline demand is one that is highly related with the price of gas and it does not affect the demand for gas. The instrumental variable that I choose is an average wage level for each of the states. Changes in wage level will influence price of supplied, but not demanded. Instrumental variable is a variable that has no influence on demand only on price. Wage is also exogenous variable because its values are coming from outside the model.

Price= real sales price of gas by state through retail outlets (average, all grades)

Real GDP= per capita real income by state (millions of chained 2000 dollars)

State sizes = square miles

Urban population = a percentage of total population (state data)

Number of cars= average number of cars per capita in each of the states

The expected signs based on the above discussion of each right-hand-side variables: β1 < 0, β2 >0, β3 >0, β4 <0, β5 >0.

3. Data Sources and Description

Table 1 Variables and descriptive statistics

Variable name Variable description Mean value Standard dev. Data Source Consumption Annual state consumption of motor gasoline 0.50 0.068 (1) per capita (thousand gallons ) Price Real price of gas ,including taxes(cents per 89.54 16.101 (2) gallon) RGDP Real GDP per capita by state (millions of 34012.42 6084.64 (3) chained 2000 dollars) State Size Size of state (square miles) 67,310.47 79559.19 (4) Urban population Urban population in each of the states as a 71.72 14.809 (5) percentage of total population Cars per person Average number of cars per person 0.45 0.092 (6) 11 Price and consumption data summary:

From the data set I used to estimate the model, prices of gas varies significantly among the states. The highest prices of gas in 2001 were in Nevada, Hawaii, New York, Oregon and Rhode Island in average of 89 cents per gallon. In 2002 each of the 50 states had experienced fall in prices of gas of average 7 cents per gallon. By 2005 the highest price per gallon was in Washington, 128 cents. There were not very big variations in per capita consumption of gas in 2001 to 2003. Noticeable change was in 2004, were all of the US states had experienced rise in per capita consumption of gas by average of 30 gallons. In 2005 there was a decrease in per capita consumption of gas in each of the US states, except Wyoming, by average of 17gallons.

Sources:

(1) US Department of Transportation, Federal Highway Administration; Office of Highway Policy Information. Highway Statistics 2000-2005 (section 1.).Monthly gasoline reported by state (total). Web page: http://www.fhwa.dot.gov/policy/ohpi/hss/hsspubs.htm

(2) Energy information administration, Gasoline prices by formulation, sales, grade type (cents per gallon excluding taxes) http://tonto.eia.doe.gov/dnav/pet/pet_pri_allmg_a_EPM0_PTC_cpgal_m.htm

(3) Bureau of Economic Analysis. Regional Economic Accounts. Real GDP by state. 2000- 2005 Web page: http://www.bea.gov/regional/gsp/action.cfm

(4) States of the United States, State size by square miles Web page: http://www.laughtergenealogy.com/bin/histprof/states/states_sz.html

(5) US Census Bureau, Rural population (resident) by state in thousand – 2001 Web page: http://www.census.gov/compendia/statab/tables/07s0033.xls

(6) Swivel preview, Data-Table: Cars per state Web page: http://www.swivel.com/data_sets/spreadsheet/1006019

12 Results

I used statistical analyses to conduct empirical analyses. A panel data set is used in estimation. Missing values (179) for some variables in the model yielded a data set with

249 observations in total. The reason for so many missing values is that my excel data include all the supporting data that I used in order to estimate per capita values (CPI, population, taxes on gas, total US consumption of gas, ext.). I estimated my model by using ordinary least squares analysis (linear and double log forms), and also two stage least squares. In my opinion results of two stage least squares explain my model better than results of ordinary least squares.

At first I estimated a linear version of equation one. Results of utilizing statistical regression analysis show that in the short run demand for gasoline is relatively price inelastic (Table 1). There is a negative relationship between price of gas and demand. As price increases 1 cent per gallon, there is a decrease in per capita consumption by 0.63 gallons a year. Also, the results are showing that there is a positive relationship between the real GDP level per capita and consumption of gas. If the real income per person goes up by

$1 there will be increase in the consumption of gas by 0.00154 gallons a year.

I also used double log to estimate the model in order to estimate percentage change in demand relative to percentage change of other variables. Results show that 43% of variation in dependant variable is predicted by independent variables. A double log form results also show that if the price goes up by 10%, demand for the gas would go down by

1.3%. Also if real income increased by 10% consumption of gas would go up by 2.1%. All of the other independent variables have the sign that I predicted it to be. From both the

13 OLS and double log results we can see that the demand is relatively price and income inelastic.

As a stated previously in the paper there is a possibility of problems with the OLS because two or more dependent variables are determined at the same time I address this by estimating the model using two stage least squares. The results show that 19% of variation in dependent variables is explained by the model. Also, β0 is 1.017 which means if all of the right hand side variables were equal zero, than per consumption of gas would be 1017 gallons per person. The results of stage one of two stages least squares show that only 8% of variation in price is predicted by wage. Rest of the results using two stage least squares show that if average annual wage increases by 1000 dollars that will cause price of gas to increase .7 cents per gallon. Price of gas and consumption of gas are inversely related.

Price variable is statistically significant at 5 percent level. If price goes up by 1 cent there will be decrease in per capita consumption of gas by 5.9 gallons a year. The results are showing that there is a positive relationship between the real GDP level per capita and consumption of gas. If the real income per person goes up by $1 there will be increase in per capita consumption of gas by 3.93 gallons a year. The size of the state and consumption of gas are also directly related. Although size of the state variable is positively related with dependant variable, it is not statistically significant. On the other hand urban population is statistically significant at the 1% level and is inversely related to consumption of gas. As the percentage of the people that live in the city increase by 1 percent, per capita consumption for gas decrease by 2.7 gallons a year. The last variable that I used for this model is number of cars per person. If average number of cars per person increases by 1, the consumption of gas will also increase by 143 gallons a year. A car per person variable is statistically significant 10% level. All of the results are consistent with my expectations.

Price and income parameter estimates are consistent with findings of earlier literature.

14 From the simultaneous equation we can say that average wage is not very good instrumental variable for this model.

4. Conclusion

This paper was very interesting and challenging journey for me. I learned couple new things that I will definitely include in doing any other projects or economic papers.

One of the most important things that I learned is that, doing economic paper is a long process of finding different data that will help you explain better your model. At the end I also came to conclusion that economic paper can never be perfect!

I am satisfied with the process and outcome of my project, however I do believe that, if I had more time, it cold be explained in more depth. Things I would change would be adding one more instrumental variable and estimating two stage least squares in double log form.

In this paper I estimate to see what factors effect consumption of gasoline in the United States, for the period from 2001-2005. Two stage least squares and both of

OLS models, linear and double log, estimated that in the short run demand for US gasoline is price and income inelastic, but two stage least squares explain my model the best If price of gas went up, demand for gas would go down but percentage change in demand is less than percentage change in price. Also, if average income per person increased, demand for gas would increase as well.

Short run price elasticity at the mean is -0.53, which confirms that demand is price inelastic. My results on price and income elasticity of demand for gasoline are very similar with elasticities estimated in Dahls paper, price elasticity is -.20. Also, results of my paper are very similar with C.J.Nicols results on short run price elasticity which is -0.162. Results

15 of my paper show that demand is less price inelastic than results of literature works I just

mention

One important question that I raised in the beginning of the paper was, whether

increase in taxes would lead to any decrease in consumption. How much would state gas

taxes have to increase, assuming full forward shifting, to reduce consumption by some

specific amount? From the two stages least squares regression analyses I can say that if

state tax on gas goes up by 2 cents (at the same time income goes up by 2 dollars), and we

assume full forward shift, per capita consumption of gas would fall by 3.86 gallons a year.

Table 2 Determinants of the gas consumption

Regressors OLS OLS Two SLS Linear Form Double log Linear form Constant 0.673*** 0.6131 1.017*** (20.32) (1.19) (5.49) Price -0.00063*** -0.138*** -.0059** (3.02) (3.47) (2.24) RGDP 0.00000154** 0.212*** 0.003933** (2.33) (3.89) (2.22) State Size 0.0655 0.032*** 0.0000402 (1.27) (4.52) (0.40) Urban population -0.00323*** -0.452*** -0.0027*** (12.24) (11.77) (4.79) Cars per person 0.132*** 0.0987*** 0.143* (2.93) (4.65) (1.65) Adj. R-Square 0.439 0.431 0.17 Observations 249 249 249 F-Statistics 39.77*** 38.57*** 11.18***

Notes:

Absolute value of t-Statistic is in parentheses

(*) Denotes statistical significance at the 10% level (**) Denotes statistical significance at the 5% level 16 (***) Denotes statistical significance at the 1% level

References:

Alan M.Schneider (1978): A New tax on gasoline: Estimating its effect on consumption. JSTOR .Science, New Series, Vol. 202, No. 4369, pp. 755-757.

Bureau of Economic Analysis. Regional Economic Accounts. Real GDP by state. 2000-2005 Web page: http://www.bea.gov/regional/gsp/action.cfm

Carol A. Dahl (August 1982). Do gasoline demand elasticities vary? Land Economics, Vol. 58, No. 3, pp. 373-382.

Energy information administration, Gasoline prices by formulation, sales, grade type (cents per gallon excluding taxes) Web page: http://tonto.eia.doe.gov/dnav/pet/pet_pri_allmg_a_EPM0_PTC_cpgal_m.htm

Hughes, Knittel, and Sperling (October 2006): Evidence of a shift in the short- run price elasticity of gasoline demand. University of California, Energy Institute

Kenneth A.Small , Kurt Van Dender( 2007). Fuel Efficiency and motor vehicle travel: The declining rebound effect. The Energy journal, Vo28, No.1. pp. 25-51.

Natural Resources Defense Council –NRDC (October 2004). Safe, Strong and Secure: Reducing America's Oil Dependence.

.Polzin S.E. and Chu, X. (2005): A Closer looks at the public transportation mode Share Trends.” Journal of Transportation and Statistics”.

Steven L. Puller, Lorna A. Greening (1999). Household adjustments to gasoline price change: an analyses using 9 years of US survey data. Energy Economics 21(1999) 27-52

States of the United States, State size by square miles Web page: http://www.laughtergenealogy.com/bin/histprof/states/states_sz.html

Swivel preview, Data-Table: Cars per state Web page: http://www.swivel.com/data_sets/spreadsheet/1006019

US Department of Transportation, Federal Highway Administration; Office of Highway Policy Information. Highway Statistics 2000-2005 (section 1.).Monthly gasoline reported by state (total). Web page: http://www.fhwa.dot.gov/policy/ohpi/hss/hsspubs.htm

US Census Bureau, Rural population (resident) by state in thousand – 2001 17 Web page: http://www.census.gov/compendia/statab/tables/07s0033.xls

18

Recommended publications