What is The Impact of Carbon Tax on Firms’ R&D Expenditure? A Quantile Regression Firm-Level Approach

What is The Impact of Carbon Tax on Firms’ R&D Expenditure?

A Quantile Regression Firm-Level Approach

ECON 490 Han Wu

Professor Catherine Douglas

1. Introduction and Motivation

It is observed that global temperature has been increased by 0.8 Celsius degree compared to it was before industry revolution in 1880 (National Research Council, 2011, p15). Many scholars have reached agreement that a 2-Celsius-degree increase in global temperature could result in a wide range of irretrievable disasters on both physical and social aspects. On the one hand, biodiversity is under the threaten of global warming because natural characteristics have been changed a lot. Therefore, habitats of many species are forced to shift and number of creatures decreases rapidly. Some species like polar bears are even at risk of extinction (Djoghlaf, 2008, WWF Global, 2016). On the other hand, since about 10% of world population is living in low-elevation coastal area, high density of population leads to high vulnerability (Meier, M. F et al., 2007). Hence, human society in low- elevation coastal area is at risk of sea level rise driven by global warming. Many studies nowadays reveal a tight connection between increase in global temperature and greenhouse gases accumulation given the observation through time (Meinshausen, M. et al. 2005). Accordingly, reducing the impact of greenhouse gases is regarded as a challenge for the following decades world widely (Hoel&Kverndokk,1996). Combustion of carbon- containing fossil fuel which is related to human activities is considered to be the main source of emission of carbon dioxide. Records of human-related emissions has been started from 1751. A dramatic increase was noticed since 1863 with the level of 0.1 GtC/year. This announced the beginning of a constant increase in emissions after industrial revolution. By the end of 1995, carbon dioxide concentration reached 6.5 GtC/year. This is a surprising increment compared to the average increment of 3% over the past 2.5 centuries (National

Research Council, 2000). Carbon content level is different across categories of energy sources, emission ranges from 13.6 MtC/EJ for natural gases, 19.0-20.3 MtC/EJ for oil to 23.9 MtC/EJ for coal (Wuebbles, D. J., & Jain, A. K. 2001). Therefore, one effective way to reduce greenhouse gases emission is to adjust energy use structure and increase investment on technology change in order to emit less carbon dioxide to the atmosphere during production. Broad consensus that we must slow down path to global warming leads to the rises of interests among governments on climate policies which design regulations on greenhouse gases emission (Stavins, 2011). As one form of carbon pricing policy which presents explicit price level, carbon tax is a commonly adopted policy which has already implemented for decades in some countries such as Norway, Finland and Denmark (world bank, 2016). By adding cost on pollution, carbon tax is considered to be a useful policy to reduce emissions of greenhouses gases because cooperatives and individuals who are covered by this policy have incentives to reduce pollution in order to pay less tax. On firm- level, cooperatives are likely to take actions such as increase investment on technology innovation which could increase efficiency and reduce emission during production (Goulder, &Schein, 2013). According to the International Panel on Climate Change (IPCC), the reduction target by 2050 is a 80% reduction in greenhouse gases emission compared to emission level in 2000. This results in higher demand on advanced technology which emits less pollution (Nicholson et al., 2011). Thus, it is reasonable to make hypothesis that carbon tax would have a positive effect on firms’ expenditure on research and development. The main objective of this paper is to analyse the impact of carbon tax on firms’ expenditure on research and development by using a quantile regression approach combine with data collected from each firms’ reports. This paper is formed as following: in

sector 2, I would introduce the background and popular debates on existed carbon policies as well as some existed research on relationship between carbon tax and R&D expenditure. In section 3, I would explain the theory which guides my empirical research. Then I will present my methodology and main findings in sector 4 and will conclude my work in sector 5.

1. Background and Literature Review
2.1 Introduction of Carbon Pricing Instruments and Debate

World Bank in 2016 reported that greenhouse gases from combustion of fossil fuel should be responsible for constantly increased global temperature, and a mass of efforts world widely is already adopted. To ensure the efficiency of climate mitigation policies, one key characteristic is that these policies, such as cap-and-trade system and carbon tax, add price on carbon-emitting activities in order to increase the cost of emitting carbon dioxide and shift incentives of emitters (Metcalf, 2007). Currently, 39 nations and 23 sub nations already participated in adding price on carbon emission. Successful implementations of carbon pricing policies can be found at different level. Some multi-countries cap-and-trade system, for instance, European Union’s Emissions Trading Scheme has been adopted since 2008 and now successfully stepped into phase 2 (but there is no implementation of multinational carbon tax right now since carbon tax rely on government while cap-and-trade scheme is a market-based scheme). On national level, cap-and-trade systems are implemented in countries like Australia and New Zealand while carbon tax is widely adopted in European and Scandinavia countries like Denmark and Norway since 1991. Sub national jurisdictions are also involved in carbon pricing instruments. For example, BC

carbon tax and a pilot cap-and-trade scheme in seven cities in china (World Bank, 2016). However, it is always a popular debate among scholars that whether cap-and-trade system or carbon tax a better climate mitigation policy. Scholar who are in favor of cap-and-trade stated that cap-and-trade system has more certainty on reduction target. Under cap-and- trade system, firms are more flexible to decide explicitly how much they would emit through purchasing or selling of the permits of emissions. This flexibility also brings extra benefits when adjusting the long-term target under the cost-and-benefit analysis framework (Stavins, 2008). On the other hand, some scholars such as members of “Pigou Club”, an academic club formed by Harvard’s Greg Mankiw who prefer carbon tax, an exogenous emissions pricing instrument, agree that carbon tax has a number of advantages over pure cap and trade (Goulde & Schein, 2013). As an inherently simple tool, carbon tax levies a clear price level on carbon content in the main sources of carbon dioxide emissions. This feature brings certainty on cost. By contrast, difficulties on setting proper baseline as total allowable emission in cap-and-trade system brings complexity for policy makers and uncertainty on cost for individuals (Stavins, 2005). Besides, carbon tax is considered to be a double dividend policy. When it is reducing emission of greenhouses gases, the revenue collected from tax payer could be recycled back to individual and cooperates for further greenhouses gases emission (Robert, 2008). For example, a paper published in 2016 by Carl and Fedor investigated revenue collected from carbon pricing instruments and pointed out that carbon tax revenue directly recycles back to tax payer as government general funds (Carl &Fedor, 2016). Moreover, carbon tax gives direct signal for producers to reduce emission during production while cap-and-trade might send ambitious signals since its reduction target can be done by trading permits. Producers might pollute as long as they are

able to pay for purchasing permits (Avi-Yonah & Uhlmann, 2009). Therefore, this paper is going to determine the impact of carbon tax instead of cap-and-trade.

2.2 Literature Review on Carbon Tax

Studies of carbon tax cover various aspects, and these studies can be generally divided into two groups based on authors’ methods. The first group is interested in determining optimal level of carbon tax and its connection to research and development by using analytical and numerical models. While the second group usually uses empirical approaches to demonstrate carbon tax’s impact in real world.

As representors of first group, Hoel and Kverndokk in 1996 combined theories of non- renewable resource and negative externalities. Based on their research, carbon stock is regarded as a non-renewable resource which should not be depleted since release of carbon stock directly results in global warming. To reduce the negative impact of greenhouse gases effectively and reserve carbon stock, amount of carbon tax should increase at the beginning and decrease after it reaches its maximum point. In the end, it should approach to zero (Hoel, M., & Kverndokk, S. 1996). Since one significant indicator to measure technology change is research and development investment, many existed researches detect relation between these two factors. Some scholars in the early time believe that carbon tax would have positive effect on R&D investment. For example, Buonanno and his collogue used three different methods to add technical change factor in an integrated assessment model applied in the case of Kyoto Agreement. After three different modifications, they concluded that carbon tax will induce research and development (Buonanno, 2003). Besides, by employing analytical and numerical models, Goulder and Mathai in 2000 also got similar conclusion. However, in 2003, Sue Wing concluded a opposite results from previous

literature that R&D investment could be negatively affected by carbon tax (Sue, 2003). Other scholars like Farzin, Kort and Baker stated their results which do not deny any of above two groups. In 2000, Farzin and Kort investigated impact of uncertain tax on firms’ investment on technology by using a dynamic model which is optimally controlled. They conclude that technology investment will increase up to a threshold level of tax. After that, investment on technology tend to decrease (Farzin and Kort, 2000). This is true that extremely high rate of tax put too much pressure on firms’ production, hence, carbon tax will lose its positive effect on investment on technology gradually (Baker, 2006).

Second group, on the other hand, often adopted empirical approaches to detect carbon tax’s impact. Bruvoll in 2004 investigated the effect of carbon tax in Norway by simulating a general equilibrium model and found that carbon tax contributed to 2.3% reduction in greenhouse gases emission (Bruvoll & Larsen, 2004). Later in 2008, Lee, Lin and Lewis analysis the combing effect of both carbon tax and cap-and-trade policies in different industry sectors. To analyse the impact of carbon tax and emission trading system on different industry sector, they use a fuzzy goal programming model and found that plastic materials and artificial fibers sectors are benefit from carbon tax and emission trading system while the accumulated GDP value in petrochemical industry has declined (Lee, Lin &Lewis. 2008). Lin and Li analysis carbon tax’s effect in European countries since carbon tax was adopted in Europe firstly and tax rate is relatively high compared to carbon tax in other countries. They employed difference in difference model and found that high rate of carbon tax combines with revenue recycling lead to effective reduction on emission (Lin &Li, 2011). Lin and Li are not the only ones who are interested in revenue of carbon tax, Carol and Fedor in 2016 investigated the current use of public revenue from national aspect

and found that 72% of revenue from cap-and-trade system and 70% of carbon tax revenue are used in green spending: government spending or subsidies on research and development on energy efficiency and renewable resources (Carl, J., & Fedor, D. 2016).

In the paper which is highly related to my work, Yu and Bi concluded that impact of carbon tax is heterogeneous based on an analysis of performance of firms in China(Yu&Bi,

2015). Their analysis gives the results that environmental tax has positive effects on firms’

abatement investment. The effect of tax is increasing as firms’ scale is increasing and decreasing after middle level. For the large-scale firm, although environmental tax also shows positive effect, the magnitude of impact is decreasing compared to the middle-level firms. In the other word, a threshold level is existed in sample firms. However, due to the fact that china is not the countries implemented environmental tax, authors use other tax which relate to environment such as resource tax, urban construction tax and land use tax as proxy indicators of environmental tax. Because of the limitation on political aspect, the results may be inconsistent if carbon tax is truly implemented in China (Bi& Yu, 2015).

After studying previous researches, it is easy to notice that most scholars are interested in the effect of carbon tax on macroeconomic level or using numerical model to detect carbon tax and its impact on R&D. Some researchers who are interested in impact of carbon pricing in microeconomics are more willing to investigate the impact of emission trade system, one other main carbon pricing policy, instead of carbon tax. Limited paper focusing on carbon tax’s impact on microeconomic level. Therefore, this paper is going to adopt Yu and Bi’s method to investigate carbon tax’s effect on firms’ expenditure on R&D.

Recently, however, several researchers have emphasized that CO2 policies and the rate of technological change are connected: to the extent that public policies affect the prices of carbon-based fuels, they affect incentives to invest in research and development (R&D) aimed at

3. Theatrical framework

This paper is based on theory of Pigouvian tax which is put forward by British economist Pigou in 1920. In this paper, carbon tax is regarded as a form of Pigouvian tax, a tax which is levied on any market activities that generates negative externalities. Carbon emission is considered to be a negative externality which adds marginal cost to the whole society. The amount of tax should be equal to the social marginal damage. Since carbon tax would increase the cost of firms’ production, this policy would shift supply curve up by the amount of tax. Firms which are taxed have incentives to reduce emission during the process of production and seek for cleaner technology and increase efficiency.

4. Data and Methodology

4.1 Variables Discussion

The main objective of this paper is to investigate if carbon tax could stimulate firms to spend more on technology innovation in order to pay less carbon tax. Therefore, 49 firms’ research and development expenditure in four different sectors is chosen as dependent variable. The main explanatory variable is TX where TX equals to zero if no carbon tax is adopted in the given country and equals to one if it is. Since carbon pricing has expand rapidly after 2012, here I would choose 2012 to 2016 as sample time period. Five control variables are total assets which measure the size of firm, number of employment which

indicates whether the firm is capital-intensive or labor intensive, profit after tax and net cash flow which measures performance of firm. However, as data for firm-level analysis usually measured in million dollars which too big compared with the unit of employee. To eliminate the difference in dimension of quantity, here I log-transferred dependent variable and other five control variables and equation is given as following:

Ln_RDS=β0+β1*TXit+β2*ln_RVit+β3*ln_AAit+β4*ln_EMPit +β5*ln_PFTit+ β6*ln_CSHit+ e

By the end of 2016, 15 countries or jurisdictions adopted carbon tax. Although some other energy taxes exist and are adopted by some countries, carbon tax has competitive mitigation effect despite different rates of tax for it concentrates ultimately on reducing carbon dioxide emission instead of energy use (Cansier and Krumm, 1997). Following table summarizes where each firm is operated and the situation of local carbon tax policy. Information of carbon tax in each countries and jurisdictions are collected from World Bank (following table only take carbon tax into consideration, other carbon pricing policies are excluded).

Table 1 country where firm is operating and local carbon tax

Table 2 and table 3 tells statistical description of variables. Dependent variable: firms’ research and development expenditure is shown in table 2. Original figures are collected form PwC’s world top 1000 research and development expenditure firms list. According to the table, total account of standard deviation is around 2.6 which means that expenditure on research and development different among firms. Viewing data year by year, it can be seen that expenditure on research and development is slightly increasing through time. No significant difference in minimum and maximum are observed in sample period. Statistical description of other control variables is shown in table 3. Firms key financial data are downloaded from DataStream. Some missing data are collected and organized from the firms’ annual report.

Table 2. Statistical description of firms’ research and development expenditure

Table 3. Other control variables’ statistical description

4.2 Model Selection

In this paper, I will use quantile regression approach to estimate coefficients of parameters in different quantiles. When we are dealing with panel data, fixed effect model with least square is one of the most commonly used model. We assume distribution of error of data satisfies Gaussian’s basic assumptions when we use least squares to estimate parameter requires data. However, in reality, many data do not reach basic assumption of least square. Existence of heteroscedasticity and non-normal distribution could sometimes result in non- BLUE (best linear unbiased estimator) results. To extend application of linear regression model, Koenker and Bassett came up with quantile regression in 1978 to estimate rates of change in different parts of the whole distribution of a specific variable by dividing into quantile. (Koenker, R., & Bassett Jr, G. 1978). Rather than least squares estimation which is insensitive to the extreme values of the data, quantile regression estimates the effect of how impact of independent variables is varying among different point of dependent variables on conditional distribution. It is a more accurate estimation when effect of independent variables near the extremes are different with the effect near the center. Since quantile regression does not have strict requirement on normality, it is especially suitable for data

which is not distributed normally and is widely adopted for analysis environment-related issue (Yu et al., 2003). Graph 1 in the following shows normality of my dependent variables used here. It is clear to see that data are not normally distributed and extreme values are existed in the right-hand side. In this paper, I would use quantile regression to exam if the impact of carbon tax varies across different quantiles on conditional distribution. It worth noticing that I also use fixed effect model and compared the results with quantile regression. Table 4 in the following shows empirical results.

Graph 1. Kdensity of dependent variables

4.4 Results and Findings

Table 4 in in the following tells empirical results about quantile regression in different quantiles and also fixed effect results. From the table, it is noticed that both quantile regression results and fixed effect results give positive relationship between carbon tax and firms’ research and development expenditure when other factors are unchanged. This result

indicates that carbon tax could stimulate technology innovation indeed by increase of firms’ production cost. To reduce cost, firms tend to increase expenditure on research and development. The impact is mainly expressed in two aspects: On the one hand, firms would adjust production process and concentrate on more efficient products under high rate of carbon tax since increasing efficiency could create higher profit and increase firms’ competitiveness in the market. On the other hand, firms are likely to prone to adopt advanced technology which emits less pollution as carbon tax is a tax which levied on the amount of emission. As advanced technology could increase efficiency and reduce emission during production, firms therefore pay less tax and reduce production cost. As a result, they could be more competitive in the market. Accordingly, carbon tax accelerates firms’ expenditure on research and development. However, it is also easy to notice that although coefficients of tax are positive from quantile 15 to quantile 80, value and significance level of coefficients are quite different from low quantile to high quantile. Value is keep increasing from q15 and reach its maximum at q60. Besides, carbon tax’s effect at low quantile level is not significant according to regression results. This could be explained by the fact that firms’ R&D expenditure are small at low quantile level, so impact of carbon tax is not significant. As firms’ R&D expenditure increasing, size of firms increasing as well. This type of firm tends to emit more. Thus, carbon tax expresses significant impact on their R&D expenditure. At high quantile level, coefficients of carbon tax are positive and significant, but value of coefficient is decreasing which indicates that high pollution firms already spend a lot on R&D expenditure. As a result, the impact of carbon tax is reducing compared to middle-level firms. To conclude, carbon tax does have positive effect on firms’ R&D investment, but impact of carbon tax is heterogeneous in different firm level.

Table 4. Quantile regression and fixed effect results

*indicated significance level. ***: p<0.01, **: p<0.05, *: p<0.1

5. Conclusion and Further Suggestion

This paper picks data of 49 firms from 2012 to 2016 as sample to analysis the impact of carbon tax on R&D expenditure. This paper started with fixed effect model, however, as data are not normal distributed, results are not significant. Therefore, quantile regression is chosen for its applicability for non-normal distributed data. Results show that carbon tax has significantly positive effect on firms’ R&D expenditure and this impact varies among different level of firms. According to results, carbon tax has positive meaning on simulating technology innovation by increasing R&D expenditure on firm level. As technology innovation level increase, emission level could reduce consequently. Besides, since governments collect revenue from carbon tax, tax revenue could recycle back to tax payer which accelerated further effort on reducing emission. Different from existed studies on carbon tax which pay attentions on macro-economic level, this paper analysis impact of carbon tax on firm level. This paper not only broadens the research field of carbon tax, but also enriches study on environmental management, which is meaning for both firm

managers and policy makers. However, due to the fact that most firms do not reveal explicit amount of carbon tax they pay, this paper simplifies the level of carbon tax each firm paid. This simplification might lead to inaccurate estimation of results. Besides, most firms do not reveal environmental R&D investment in their annual report. this paper uses general R&D expenditure because environmental R&D is proportional to general R&D. This might also cause inconsistent estimation with true model. Further research direction could be using questionnaire to investigate explicit amount of firms’ carbon tax payment and environmental research and development investment to get more accurate estimation.

Reference List:

Avi-Yonah, R. S., & Uhlmann, D. M. (2009). Combating global climate change: Why a carbon tax is a better response to global warming than cap and trade. Retrieved from: https://repository.law.umich.edu/cgi/viewcontent.cgi?article=1051&context=art icles

Bruvoll, A., & Larsen, B. M. (2004). Greenhouse gas emissions in Norway: do carbon taxes work? Energy policy, 32(4), 493-505.

Buonanno, P., Carraro, C., & Galeotti, M. (2003). Endogenous induced technical change and the costs of Kyoto. Resource and Energy economics, 25(1), 11-34.

Cansier, D., & Krumm, R. (1997). Air pollutant taxation: an empirical survey. Ecological economics, 23(1), 59-70.

Carl, J., & Fedor, D. (2016). Tracking global carbon revenues: A survey of carbon taxes versus cap-and-trade in the real world. Energy Policy, 96, 50-77

Djoghlaf, A. (2007). Climate Change and Biodiversity in Polar Regions. Sustainable Dev. L. & Pol’y, 8, 14

Farzin, Y. H., & Kort, P. M. (2000). Pollution abatement investment when environmental regulation is uncertain. Journal of Public Economic Theory, 2(2), 183-212.

Goulder, L. H., & Mathai, K. (2000). Optimal CO2 abatement in the presence of induced technological change. Journal of Environmental Economics and

Management, 39(1), 1-38.
Goulder, L. H., & Schein, A. R. (2013). Carbon taxes versus cap and trade: a critical

review. Climate Change Economics, 4(03), 1350010.

Hoel, M., & Kverndokk, S. (1996). Depletion of fossil fuels and the impacts of global warming. Resource and energy economics, 18(2), 115-136.

IPCC (Intergovernmental Panel on Climate Change. Climate change 1995—economic and social dimensions of climate change, vol. 3, Cambridge University Press, Cambridge (1996)

Koenker, R., & Bassett Jr, G. (1978). Regression quantiles. Econometrica: journal of the Econometric Society, 33-50.

Lee, C. F., Lin, S. J., & Lewis, C. (2008). Analysis of the impacts of combining carbon taxation and emission trading on different industry sectors. Energy Policy, 36(2), 722- 729.

Lin, B., & Li, X. (2011). The effect of carbon tax on per capita CO 2 emissions. Energy policy, 39(9), 5137-5146.

Meier, M. F., Dyurgerov, M. B., Rick, U. K., O’Neel, S., Pfeffer, W. T., Anderson, R. S., … & Glazovsky, A. F. (2007). Glaciers dominate eustatic sea-level rise in the 21st century. Science, 317(5841), 1064-1067.

Meinshausen, M. et al. (2009). Greenhouse gas emission targets for limiting global warming to 2 °C. Nature 458, 1158–1162

Metcalf, G. (2007). An equitable tax reform to address global climate change. Retrieved from: https://www.brookings.edu/research/an-equitable-tax-reform-to-address-global- climate-change/

National Research Council. Reconciling Observations of Global Temperature Change, National Academy Press, Washington, DC (2000)

Nicholson, M., Biegler, T., & Brook, B. W. (2011). How carbon pricing changes the relative competitiveness of low-carbon baseload generating technologies. Energy, 36(1), 305-313.

Robert W. Hahn, Greenhouse Gas Auctions and Taxes: Some Practical Considera-tions (Am. Enter. Inst., Ctr. for Reg. & Mkt. Studies, Working Paper No. 08-12, (2008).

Schleich, J., & Betz, R. (2005). Incentives for energy efficiency and innovation in the European Emission Trading System. In Proceedings of the 2005 eceee Summer Study— What Works & Who Delivers.

Stavins, R. N. (2008). A meaningful US cap-and-trade system to address climate change.

Retrieved from: http://www.law.harvard.edu/students/orgs/elr/vol32_2/Stavins%20Final%20Fin al.pdf

Stavins, R.N. (2011). The problem of the commons: still unsettled after 100 years. Am. Econ. Rev., 101 (2011), pp. 81-108, 10.1257/aer.101.1.81

World bank Economics, V. (2016). State and Trends of Carbon Pricing 2016. World Bank. Retrieved from:

http://documents.worldbank.org/curated/en/598811476464765822/State-and- trends-of-carbon-pricing

WWF Global (2016). WWF Global: Arctic climate change. Retrieved from: http://wwf.panda.org/what_we_do/where_we_work/arctic/what_we_do/climate /

Wuebbles, D. J., & Jain, A. K. (2001). Concerns about climate change and the role of fossil fuel use. Fuel Processing Technology, 71(1), 99-119.

Yu, K., Lu, Z., & Stander, J. (2003). Quantile regression: applications and current research areas. Journal of the Royal Statistical Society: Series D (The

Statistician), 52(3), 331-350.
Yu, L., & Bi, Q. (2015). Relationship Between Environmental Taxes and Enterprise Green

Investment Behavior: A Panel Quantile Regression Approach. Chinese population, resource and environment,2016,26(03):76-82.