# «VOLUM E 1 1, N UM B E R 1 I S SN 2 1 6 8 - 0 6 1 2 F L ASH DR I V E I S SN 1 9 4 1 - 9 5 8 9 ON L I N E T h e In s t it ut e f o r Bu s i n e s s an ...»

are influenced by gasoline price( ), interest rate(( )), wealth( ) and energy efficient( ).

Equation (14) is gasoline demand unrelated to TCC. Equation (15) is gasoline demand related to TCC. Both However, gasoline demand in equation (15) is affected by TCC such as interaction term( ( )) and a direct impact on TCC.

Regional Gasoline Demand Function This paper estimates regional gasoline demand in South Korea. Equation (14)-(15) derive by a dynamic optimization is generalized gasoline demand function. we let their functions to take logarithm and to rearrange. They are expressed by equation (16) and equation (17) , = + 1 ln , + 2 , + 3 + 4 , + , (16)

where i is region in south Korea, t is time, , is gasoline consumption per capita in region, , is average gasoline price at gas station in region, , is GRDP(Gross regional domestic product) per capita, , is energy efficient, is interest rates as proxy variables of rental price, , is TCC, , is a contemporaneous error term. In equation (16), coefficient of gasoline price, interest rate, and energy efficient shows negative signal, income is positive(Batagi and Griffin, 1988; Medlock and Soligo, 2002).

In equation (17) related by TCC, expected on the coefficient of Income, interest rates and energy efficiency is the same as in equation (16). But price elasticity influenced by cross-term gasoline can be expressed as equation (18).

We expects 1 + 2 , 0, since gasoline price have negative(-) relationship on gasoline demand.

cross-term between price and TCC has negative relationship with gasoline demand, so that 2 0 is Futhermore, The gasoline price elasticity varies with the TCC in equation (18). According to equation (15), transport. It is expected to have a negative sign of the Traffic congestion cost elasticity (1 + 2 , ) also expected. It is implied that increasing TCC increases magnitude of price elasticity due to use of public by the same reason.

Empirical Methodology In order to estimating equation (16) and (17), There are potential econometric issues about endogeneity.

Estimating demand function occurs because price and consumption are jointly determined by the interaction of supply and demand curves(Lin and Zeng, 2013). This endogeneity will leads to a biased parameter estimates. Through DWH(Durbin-Wu-Hausman) Test, we will check the endogeneity of gasoline price empirically. An instrument should satisfy the two condition: it should be highly correlated with gasoline price and should not be correlated with, contemporaneous error term. Batagi and Griffin(1997) used the lagged exogenous variables, Ramsey et al(1975) used gasoline inventory quantity, Lin and Zeng(2013) used diesel prices as instrumental variables. we use also exogenous variables, gasoline inventory quantity and diesel prices as instrumental variables. and see two conditions about instrumental variables using some test. For test instrumental variables unrelated with error term, we use Hansen J statistic. In order to strength of the instrumental variables, we use AR(Anderson-Rubin F Test). TSLS(Two Stage Least Square) and GMM(Generalized Method of Moments) is representative estimator as instrumental variables method. This study compares the estimated model with 2SLS and GMM, and the results of the regression equation and the fixed effects model was analyzed.

## DATA AND RESULTS

Data The data are panel data which range from 2005- 2012 for each region in South Korea. Gasoline consumption per capita (, ) obtained from the petronet (www.pertronet.com) and Korea Statistical information service.Since gasoline consumption presented in Perronet is BbL basis, we will convert to liter. Gasoline price(, ) is also regional gas station price data from Petronet, Interest rate( ) and Income(, ) obtained from Korea Statistical information service. Finally, TCC(Traffic congestion cost (, ) per capita is from Korea Transport Database(www.ktdb.go.kr). All variables was adjusted to real variables through the regional efficient, , obtained from Survey of distance travelled by vehicle (Korea Transportation Safety Authority) CPI(The regional consumer price index in South Korea) from Korea Statistical information service. Energy from 2005 to 2012. Descriptive statistics of the variables are as follows Table 1.

Table 1: Descriptive Statistics

## RESULTS

This paper estimate gasoline demand function using equation (16) and (17) The results of equation (16) which is disregarded TCC are in Table 2, where column (1) is results of OLS, column (2) is Fixed effect, column (3) is 2SLS and column (4) is GMM. We test presence of the endogeneity about price as DurbinWu-Hausman test and condition of ideal instrumental variables using Hansen J test and Anderson-Rubin test. The results of test is also reported in Table 2. According to results of Durbin-Wu-Hausman test, ‘gasoline price is exogenous variables.’ is rejected at 10% significance level. Therefore, gasoline price variables empirically have the endogeneity problem. We find out our Instrumental Variable is unrelated by error term using Hansen J statistics. The results follows that ‘Instrumental Variable is exogenous variables.’ GCBF ♦ Vol. 11 ♦ No. 1 ♦ 2016 ♦ ISSN 1941-9589 ONLINE & ISSN 2168-0612 USB Flash Drive 437 Global Conference on Business and Finance Proceedings ♦ Volume 11 ♦ Number 1 is mot rejected at 10% significance level. The results of test the strength of the instrumental variables such as Anderson-Rubin F shows that the null hypothesis which does not imply a correlation between the instrumental variables and endogenous variables were rejected at 1% significance level. Above results, OLS leads to a biased parameter estimate and our instrumental Variables are empirically appropriate. Therefore, 2SLS and GMM was more reliable than OLS. FE. The results of estimates equation (16) in Table 2 follows that, the estimates of coefficients about the main variable was consistent with expectations. Under OLS, 2SLS, GMM, coefficients of gasoline price, energy efficient, interest rate show negative(-) sign, and Income is positive(+) at 1% significance level. However, FE(Fixed effect) estimator shows that Income is negative(and gasoline price was not significant.Table 2: the Mode without Holding Traffic Congestion Costs

The focus of this study is to analyze the effect of traffic congestion costs on gasoline demand and price elasticity. This study estimates the effect of traffic congestion costs on gasoline demand at equation (17), and examine model specific test, test of the endogeneity, and condition of instrumental variabels. All resluts are reported in Table 3. According to results of Durbin-Wu-Hausman test, ‘gasoline price is exogenous variables.’ is rejected at 5% significance level. Therefore, gasoline price variables empirically have the endogeneity problem. In order to check that our instrumental variables are empirically appropriate, we use Hansen J statistics as test bad IV and Anderson-Rubin F as test the strength of the instrumental variables.

The results of test implied that Our instrumental variables are exogenous variables and not weak Instrumental Variable. It implies that 2SLS and GMM was more reliable than OLS, FE.

GCBF ♦ Vol. 11 ♦ No. 1 ♦ 2016 ♦ ISSN 1941-9589 ONLINE & ISSN 2168-0612 USB Flash Drive 438 Global Conference on Business and Finance Proceedings ♦ Volume 11 ♦ Number 1 Table 3: The Mode with Holding Traffic Congestion Costs

note: ***p0.01, **p0.05, *p0.1 DWH is Durbin-Wu-Hausman. AR F test is Anderson-Rubin F test, SW is Stock-Wright test The results of estimates equation (17) in table 2 follows that, coefficients of energy efficient, interest rate show negative(-) sign, and Income is positive(+) at 1% significance level, except to FE(Fixed effect).

FE(Fixed effect) estimator shows that Income is negative(-) and gasoline price was not significant. second, price elasticity are significant at l% level. To analyze price elasticity, we show 1 + 2 ,, which is estimates of price elasticity in OLS and FE are not significant. However, under 2SLS, GMM, estimates of price elasticity in equation (17). It means that magnitude of price elasticity is changed with degree of TCC.

The results is reported in table 4. For the Min value of Traffic congestion, The OLS and FE estimates results are positive sign. However, 2SLS, GMM estimates results are negative(–0.564~–0.792) with all interval of TCC, For 2SLS and GMM, Following results such as test of endogeneity and test of instrumental variables, 2SLS, GMM results are more empirically appropriate than OLS, FE’s.

Table 4: Price Elasticity ‘1 + 2 , ’

The next issue concerns the difference in price elasticity and TCC elasticity between Metropolitan and NonMetropolitan area in South Korea. Table 5 is reported that price elasticity and TCC elasticity is estimated under GMM in Table 3. results follows that: first, TCC elasticity is similar between Metropolitan(-0.198) and Non-Metropolitan area(-0.197). Because there are no difference in gasoline price across region However, we see difference in price elasticity between Metropolitan(-0.739) and Non-Metropolitan area(with TCC. It implies that as TCC increase the price elasticity increase. However, we shows that price elasticity is less than 1 under both Metropolitan and Non-Metropolitan area. It is different about results of table 2. The findings imply that price elasticity in the model without holding traffic congestion costs were overestimated otherwise and considering congestion costs may improve estimates and predictions of gasoline demand.

Table 5: Price and TCC Elasticity across Region

## CONCLUSIONS

This paper examined the effect of traffic congestion costs on gasoline demand. For analysis, we utilize gasoline demand function derived by a dynamic optimization. Allowing that distance per vehicle is influenced by traffic congestion costs, we analyze gasoline demand with holding traffic congestion costs.In addition, we deal with endogenity problem of gasoline price using 2SLS, GMM and compare their results to OLS’s. The results follows that first, the estimated coefficients were as expected. Coefficient of price, energy efficient, interest rate and congestion cost shows negative sign and coefficient of income is positive sign. They are also significant. Second, A Model specific test rejects model without holding traffic congestion costs. It means that traffic congestion cost has important role of estimating gasoline demand.

Third, we show that price is endogenous variables and our instrumental variables are empirically appropriate. Therefore, 2SLS and GMM was more reliable than OLS. Finally, he higher traffic congestion costs were associated with the higher price elasticity. While consumers in the Metropolitan area confront high traffic congestion costs, they seem to react flexible on price changes because they can use other choices of transportation (bus, subway, etc.) except for their own cars. In addition, price elasticity is less than 1. It means that price elasticity in the model without holding traffic congestion costs were overestimated otherwise. The findings imply that considering endogeneity of price and congestion costs may improve estimates and predictions of gasoline demand.

## REFERENCES

Minsung Kim and Sungsu Kim, (2011) “Estimator of Price and income elasticities on gasoline demand and diesel demand”, Journal of Environment in Korea Vol 50, p. 159~182.SuKwan Jung, (2015) “Empirical Research on Regional Gasoline Demand Efficiency”, Journal of the Korean Regional Development Association, Vol 27, p. 125~142.

Hanseon Cho and Dongmin Lee, (2008) “2007 Traffic Congestion Costs: Estimation and Trend Analysis”, The Korea transport institute,.

Anderson, T. W., and H. Rubin. (1949) “Estimators of the Parameters of a Single Equation in a Complete Set of Stochastic Equations.” The Annals of Mathematical Statistics, Vol 21, p. 570~582.

Baltagi, B. H., and J.s M. Griffin, (1983) “Gasoline demand in the OECD: An application of pooling and testing procedures,” European Economic Review, Vol 22, pp. 117~137.

Baltagi, B. H., and J. M. Griffin, (1997) “Pooled estimators vs. their heterogeneous counterparts in the context of dynamic demand for gasoline,” Journal of Economics, Vol 77, pp 303~327.

Espey, M., (1998) “Gasoline demand revisited: an international meta-analysis of elasticities,”Energy Economics, Vol 20,. 273~295.