An Empirical Assessment of the Effectiveness of Green Policies

Transcription

1 An Empirical Assessment of the Effectiveness of Green Policies Julian Dieler a, Darko Jus b and Markus Zimmer a a ifo Institute for Economic Research, Munich b Center for Economic Studies (CES), Ludwig-Maximilians-University, Munich Abstract Our empirical analysis covers a subset of the OECD-countries for the years With this data we make two major contributions existing literature on estimating oil demand functions: First, we identify the effect of green policy measures on the demand for oil. This was not yet done in any empirical study in this comprehensive way. Second, we do this on an international scale. We follow the empirical strategy of Davis and Kilian (2011) who compare different econometric methods and argue in favor of an instrumental variable approach for estimating the effect of a gasoline tax on carbon emissions in the United States. Similarly we estimate tax elasticities of oil demand for our multi-country sample. Thereby our approach differs from other studies which estimate the price elasticities only. We find that the argument of a pigouvian tax can be supported and that the influence of taxes and subsidies differ substantially between the analyzed countries. Our analysis emphasizes that green policy measures are increasingly used in many countries. Usually they are justified by the idea to reduce climate change. Our paper contributes to answering the question of how well this aim is being achieved and to which extent carbon taxes simply raise revenues. ifo Institute for Economic Research, Poschingerstr. 5, Munich; Phone: +49/89/ , Fax: +49/89/985369, dieler@ifo.de 1

2 1 Introduction The ongoing public debate about the most efficient and applicable policy interventions to reduce the speed of global warming deserves empirical examination in order to evaluate the effectiveness of the discussed policy instruments. From a practical point of view many scientists will argue that all polluting nations have to reduce their demand for oil and other carbon resources in order to emit less of the climate-damaging carbon dioxide gas. This seems to be reasonable as in fact only a few countries in the world are suppliers of carbon resources, so that the set of policy options for most countries is limited to demand side interventions. The internationally most implemented and perhaps politically most feasible policy instruments can be summarized under the idea of pigouvian taxation, i.e. the taxation of the consumption of fossil fuels and their complements. Following the Pigouvian argument environmental taxes do not have a fiscal motivation but are primarily used to correct for market failures, e.g. negative externalities. These may arise from individual overconsumption of carbon resources as the climate can be regarded as a global public good with initially a price of zero. Therefore, most countries in the world impose a price by setting taxes. This preliminary version of our paper explicitly aims at examining the effects of taxes which are aiming at a reduction of the consumption of oil only. But we are already working on the database which enables us to analyze the other two carbon containing energy products (coal and gas) and further green policy measures like subsidies. In addition to the intention of reducing the traded quantity of oil, there are, however, also reasons why taxation of oil demand (and a shift in the demand curve) might be welfare enhancing from the perspective of oil consuming countries. Bergstrom (1982) has argued that a sales tax on oil would shift rents from producers to consumers if the supply of oil was too a large degree inelastic. Then the incidence of such a tax would fall entirely on the suppliers while tax revenues would be generated in the consumer countries. As we do not consider the supply side in this paper, this conclusion remains perfectly valid if we assume the oil supply to be inelastic. In contrast if it was elastic and the demand would react to a tax on oil, then the traded quantity of oil would certainly change as well while the net price would not fall for the full amount of the tax. In any case, the effect and the interpretation of a tax on the demand remains the same. Another argument that is to test is that oil is taxed highly because of its demand being inelastic. In this case optimal taxation theory would suggest to tax inelastic goods higher than elastic ones in order to minimize welfare losses. Thus if revenue generation is the main effect of oil-taxation we should not observe significant negative effects on the demand. 2

3 In this paper we present OLS type and instrumental variables estimates of an oil demand function for a subset of the OECD-countries. First we estimate price elasticities to check if our results are in line with those of the existing literature as most focus on price elasticities. In a next step we use an IV approach to cope with the endogeneity of the oilprice in a demand function. According to Lucas W. Davis (2011) we use the change in taxes as an instrument. In Section 2 we explain our data set and further plans of extending it. Afterwards we derive from a simple model the demand function we estimate. In Section 4 we present our results, before we conclude in Section 5. 2 Data The data for our empirical analysis covers the years from 1965 to 2010 on a yearly basis. To control for the economic phase and performance of the analyzed countries we include GDP per capita. Additionally we control for the prices of coal, oil and gas to be able to distinguish between price effects on demand and the effects of environmentally related taxes in which we are mainly interested in. The tax data mainly stem from the OECD Revenue Statistics (OECD, 2012) and the IEA Energy Prices and Taxes Statistics database (IEA, 2012a). The former database contains revenues from all tax bases for the 34 OECD countries. The latter contains energy end-use prices and pre-tax prices from which implicit tax rates can be calculated. As we are interested in the effects of taxes on the emission of climate gases we select environmentally related taxes. We do so following the definition of Eurostat (2001). According to their definition environmentally related taxes are taxes whose tax base is a physical unit (or a proxy of it) of something that has a proven, specific negative impact on the environment (Eurostat, 2001). This definition comprises environmental taxes which ought to correct the externality of environmental damages (Pigouvian taxes) as well as fiscally motivated taxes whose purpose it is to collect money for the state. Oftentimes environmentally related taxes cannot be separated by the concept of environmental and fiscal taxes, as many of them shall serve both aims. Bruvoll (2009) tries to identify the two parts in the environmentally related taxes of Norway. As we want to assess the demand effects of both, the pigouvian part and the fiscal part of the taxes, we take all environmentally related taxes into account in our empirical analysis. It should additionally be noted that ad valorem taxes are excluded from the definition of the environmentally related taxes (Eurostat, 2001). This suits us very nicely as otherwise the exogeneity assumption for the regressors would not hold any longer while regressing demand of energy products on the taxes imposed on them. 3

4 Using the (OECD, 2012) database on tax revenues and the energy consumption data from another OECD database (IEA, 2012b) we calculate two implicit unit tax indexes for each energy source (coal, oil and gas). One index for those taxes which are explicitly taxing the consumption of energy products and the other one for those taxes which tax complementary goods and services (e.g. vehicle taxes or road duties). With the data we obtained in the first step of the project from the OECD and IEA databases we construct a panel, which allows us to conduct aggregate estimations (over all countries), panel estimations and country level estimations for 8 countries for oil products. In the final state the data on the climate change relevant policy measures will be used to construct a panel of implicit policy indexes considering all three carbon containing energy sources for all OECD countries plus the 4 most relevant emerging countries with respect to climate change. However the analysis of the first version is limited to the analysis of the effectiveness of the oil product taxation for those countries with good data availability. In the following we briefly explain the tax revenue paths and tax revenue drivers in each of these countries. Figure 1: Tax rates and demand for Denmark and Finland The Scandinavian countries (see Figure 1) were the precursors in taxing CO 2. Finland already implemented a tax of 90 US-Dollars per ton of emitted carbon dioxide in The Danish government followed the Finnish only one year later in 1991 when energy taxes were reformed into an energy tax and a carbon tax. These two countries were also among the first considering environmental tax reforms (ETR). Denmark introduced 4

5 their ETR in 1994 and Finland in Thereby was Finland one year after the United Kingdom who already 1996 replaced parts of the social security contributions by taxes on landfill (Bosquet, 2000). The British government imposed a fuel duty a fuel duty escalator in 1993 which basically introduced a fuel tax. Several tax increases between 1993 and 1999 let British fuel price rise from one of the lowest to one of the highest levels within the EU. The portion of the taxes in the end-use fuel price went up to 75 percent. However, due to to emerging protests the government abandoned the fuel escalator in 2000 as can be seen in Figure 2. Since 2007 the government again commenced to gradually increase fuel taxes. Figure 2: Tax rates and demand for United Kingdom and the Republic of Ireland The neighboring Irish government presented its Green Paper on Sustainable Energy suggesting a carbon tax in line with an emission trading system in Currently a carbon tax of 15 e per ton of CO 2 is charged. The German government contemporaneously presented its Eco-Tax Reform in 1999 essentially attempting to reduce fossil fuel consumption. The reform had a three-stage design with further tax increases scheduled for 2000 and 2003 and a constant tax thereafter. In addition a National Climate Change Protection Program had been enforced in 2000 which among other measures included an additional levy on air traffic. These objectives are currently via the Integrated Climate Change and Energy Program which aims to reduce greenhouse gas emissions by 40 percent until 2020 compared to the 1990 level. In order to to fulfill its Kyoto targets Switzerland implemented a carbon tax per ton 5

6 in 2008 and the government agreed to gradually increase carbon taxes according to a three-step plan when it becomes foreseeable that emission reductions are not sufficient to reach the climate goals. Figure 3: Tax rates and demand for Germany and Switzerland Figure 4: Tax rates and demand for Japan and Portugal Japan realized an Energy Tax Reform on Fossil Fuels in 2003 including a revision of the tax on fossil which, however, remained unchanged. Moreover the government induced 6

7 a preferential taxation for fuel-efficient cars in order to encourage the development of hybrid and electric cars. Concerning environmentally related taxes Portugal relies mainly on the taxation of the transport sector via fuel and vehicle taxes. And although the total revenues from environmentally related taxes decreased over the past years the implicit tax rates for oil products follow an upward sloping trend similar to all the other analyzed countries. 3 Empirical Strategy Our empirical strategy bases on the empirical approach of Lucas W. Davis (2011). They analyze the effect of gasoline taxes on carbon emissions in the United States by means of a sound set of empirical methods. They have got monthly gasoline price, tax and consumption data for the 51 US states. To compare their results with those from the existing literature they start estimating the price elasticity of gasoline demand with standard single-equation regression models using OLS and instrumental variables estimation methods. To control for time fixed demand patterns they estimate the OLS and IV regressions also in a panel setting. As a further method to cope with the price endogeneity they suggest a recursively identified vector autoregression which allows them to identify the price and the tax effect on demand. In this preliminary version of our paper we place our focus on the OLS and IV approach in single-equation and panel settings. The theoretical foundation of the population model, we estimate, is as follows: Oil is used as production factor within an economy. The typical assumptions about the production technology being Cobb-Douglas or CES will result in isoelastic demand curves. Log-linearization of this demand curve allows the coefficients to be directly interpreted as elasticities. The estimated price elasticities quantify the percentage change of oil demand for a given one percent increase in the price. We assume small economies that do not influence the oil price with their oil demand. Following Cooper s (2003) estimation of price elasticities of demand we employ an adaptation of Nerlove s (1958) partial adjustment model. Contrary to Cooper we use growth rates for the estimation. Further discussion of the approach can be e.g. found in Kennan (1979), Shapiro (1986) or in Blundell s (1988) survey paper. Following the theoretical considerations we define the demand function to be: q t = e β y γ t pπ t (1) With e β being a constant, q the per capita demand for oil products, y the real GDP 7

8 per capita in national currency and p the real price of automotive diesel. The estimation equation is specified as first differences of the log-linearized demand functions. ( ) ( ) ( ) qt yt pt ln = β β + γ ln + π ln q t 1 y t 1 p t 1 Equation (2) is more intuitive to read if you assume that the logs are an approximation of the growth rates ˆx t ln(x t /x t 1 ). This leads us to the following equation. (2) ˆq t = γ ŷ t + π ˆp t (3) The main focus of this analysis is placed on the total price effect on demand. Explicitly, as we can observe above, π can directly be interpreted as the price elasticity of oil demand. This follows directly from the elasticities for the assumed demand function: With ε q,π as the price elasticity of oil demand. ε q,π = q t π t πt q t (4) In the instrumental variables approach we instrument, according to Lucas W. Davis (2011), for the oil price by using the natural logarithms of the changes in oil unit taxes. 4 Results With the help of the OLS estimation method we estimate in a first step price elasticities of oil demand. As estimating the demand with price as one of of independent variables leads to the problem of endogeneity the estimates of the price elasticity are overestimated. Therefore we estimate in a second step tax elasticities of oil demand using an IV approach. 4.1 Least Squares Estimates Figure 5 shows the results of the single equation regressions for those eight countries with the best data availability in two specifications. The price elasticities we find in the specification in which we control for GDP growth vary between (Portugal) and (Ireland). The results of the regression which does not control for GDP growth should be interpreted with precaution as they turn out to be a bit too large. We estimated this regression, however, to attain significant estimates of the coefficients. Further improvement of our database hopefully allows us in the near future to conduct estimations in 8

9 a symmetric specification and for more than these eight countries. Furthermore we are aware of the fact that leaving out control variables for other demand influencing factors might lead to a left out variable problem. To control for time fixed effects we conduct an estimation in a panel setting. Country OLS Estimates Oil demand Denmark Finland Germany Ireland Oilprice (0.109)*** (0.031)*** (0.086)*** (0.114)*** GDP (0.055)*** Const (0.977) (0.198)*** (0.574)*** (0.972) R N * p < 0.1; ** p < 0.05; *** p < 0.01 Oil demand Japan Portugal Switzerland United Kingdom Oilprice (0.116)*** (0.065)*** (0.133)*** (0.042)*** GDP (0.937) (0.052)*** Const (1.935)* (0.665)*** (0.977) (0.268)*** R N * p < 0.1; ** p < 0.05; *** p < 0.01 Figure 5: Results from the OLS regression of country data Figure 6 shows the result of this panel estimation. The price elasticity of reflects an average over all analyzed countries. With this result we are in line with those of Lucas W. Davis. Although the panel approach is less prone to the price endogeneity we further address this problem in the next section by using an instrumental variables 9

10 approach. Panel OLS Estimates Oil demand Panel (38 countries) Oilprice (0.027)*** GDP (0.052)*** Const. R N 771 * p < 0.1; ** p < 0.05; *** p < 0.01 Figure 6: Results from the panel OLS regression 4.2 Instrumental Variables Estimates To estimate the national diesel demand tax elasticities we follow the approach of Davis and Kilian (2011). We instrument the automotive diesel price using inflation-adjusted changes in the log of the tax per ton of oil equivalent 1. The taxes are extracted from the IEA Energy Prices and Taxes Statistics database. Figure 7 reports the results of the IV estimates. Our estimated tax elasticities lie between for the Polish industry and for the Austrian industry. For the households the respective values lie between for Sweden and for Austria. These figures are in range with estimates in other recent studies focusing on gasoline taxation in the United States. In the specification favored by Davis and Kilian the IV estimate of the tax elasticity is and the baseline estimate of Bento et al. (2009) is equal to In general the elasticity of the industry seems to be higher than the elasticity of households. The lower elasticity of households might also be the result of a substitution effect. Increases in diesel taxes a typically accompanied by increases in gasoline taxes and as both increase there is a substitution from gasoline to diesel observable. This effect is less dominant in industry since this sector already dominantly uses diesel. 1 Davis and Kilian use tax per gallon. 10

11 IV estimates of diesel demand Variables Austria Austria Chile Germany Germany households industry households (1) households industry Diesel tax (0.125)** (0.245)* (0.204) (0.061)*** (0.183) Const (0.010)*** (0.010)** (0.024)* (0.005)*** (0.007)** N * p < 0.1; ** p < 0.05; *** p < 0.01 (1) no observations for Chile industry Variables Ireland Ireland Netherlands Netherlands Norway Norway households industry households industry households industry Diesel tax (0.122)* (0.197)* (0.049) (0.283) (0.093)* (0.140) Const (0.013)*** (0.016)*** (0.005)** (0.014) (0.008) (0.012) N * p < 0.1; ** p < 0.05; *** p < 0.01 Variables Poland Poland Sweden Sweden Switzerland Switzerland households industry households industry households industry Diesel tax (0.202) (0.037)** (0.035)*** (0.050) (0.154)* (0.389) Const (0.026)** (0.017)*** (0.005)*** (0.006)** (0.007)*** (0.011) N * p < 0.1; ** p < 0.05; *** p < 0.01 Figure 7: Results from the instrumental variables regression 11

12 5 Summary and Outlook In this preliminary version of our paper we mainly show that we are able to produce results with our dataset which are in the range of those in the existing literature. And we do so with international data and not merely with national datasets how it is mainly done so far. By broadening our database we will be able to provide policy comparisons within countries as well as between countries in order to ex-post evaluate the effectiveness of measures and ex-ante derive policy recommendation. For the assessment of green policies we will also include other carbon containing energy products (coal and gas) as well as other policy measures like subsidies. Furthermore we will analyze the historical development of the effectiveness of green policies by divide our sample in several time horizons. Regarding the econometric method we will also adopt a vector auto regression estimator on our data and conduct several robustness checks. References Bento, Antonio M., Lawrence H. Goulder, Mark R. Jacobsen and Roger H. von Haefen (2009), Distributional and efficiency impacts of increased u.s. gasoline taxes, American Economic Review 99, Bergstrom, Theodore C. (1982), On capturing oil rents with a national excise tax, American Economic Review 72(1), Blundell, Richard (1988), Consumer behaviour: Theory and empirical evidence-a survey, The Economic Journal 98(389), Bosquet, Benoit (2000), Environmental tax reform: does it work? empirical evidence, Ecological Economics 34, a survey of the Bruvoll, Annegrete (2009), On the measurement of environmental taxes, Research Department of Statistics Norway Discussion Papers 599. Cooper, John C.B. (2003), Price elasticity of demand for crude oil: estimates for 23 countries, OPEC Review pp Eurostat (2001), Environmental taxes - A statistical guide, European Communities, Luxembourg: Office for Official Publications of the European Communities,. 12

13 IEA (2012a), End-use prices: Energy prices in national currency per toe, IEA Energy Prices and Taxes Statistics (database). IEA (2012b), World energy balances, IEA World Energy Statistics and Balances (database). Kennan, John (1979), The estimation of partial adjustment models with rational expectations, Econometrica 47(6), Lucas W. Davis, Lutz Kilian (2011), Estimating the effect of a gasoline tax on carbon emissions, Journal of Applied Econometrics 26, Nerlove, Marc (1958), Adaptive expectations and cobweb phenomena, The Quarterly Journal of Economics 72(2), OECD (2012), Oecd tax statistics - comparative tables. Shapiro, Matthew D. (1986), The dynamic demand for capital and labor, The Quarterly Journal of Economics 101(3),