Simulating the Regional Effects of a Carbon Tax: An Application of the NEMS Energy Model Linked with the REMI Macroeconomic Model*

Frederick Treyz, Sherri Lawrence, Sinan Hastorun** (fred@remi.com)

I.   Introduction

This paper demonstrates how the integration of the NEMS-Global Insight (referred to hereafter as NEMS-GI) Model with the REMI Policy Insight (PI+) Model allows policymakers, private industry, and researchers to evaluate on a regional basis and over time the macroeconomic effects of energy and environmental policy decisions.  The current macroeconomic module of the NEMS model is a top down Global Insight model, that is, macro-economic changes are predicted on the national level and then shared-out proportionally to the regional level.  By combining NEMS and REMI, the integrated system allows for detailed energy information to drive changes in the macro-economy using a bottom-up multiregional economic model.  We demonstrate the performance of NEMS-REMI in a case study of a simulated carbon tax, and compare the results to the NEMS-GI output.

The current top-down macroeconomic capability of NEMS-GI is extended in two ways by its integration with REMI PI+.  First, the top-down approach may overly simplify the way that region-specific changes in energy prices and demand occur differently across the U.S.  In the current top-down approach, these changes are aggregated and drive a single-region national economic model, after which the changes in macroeconomic activity are shared out proportionally to the Census Divisions of the U.S.  Using the REMI multi-regional model, the energy price and demand changes drive macro-economic changes at regional level, including interactions among regions.  For example, different changes in electricity prices for the Pacific and Southwest regions are reflected in policy variable changes in the REMI framework, and their effect on the regional economy and interaction with other regions are captured in the multi-regional model.  The second extension provided by the NEMS-REMI integration is the potential application of NEMS-GI at a more geographically-detailed level.  As the REMI PI+ model can be built for any state, county, or combination of thereof, the NEMS-REMI system can show macroeconomic effects of policies for any region that may be of interest.

The paper begins by describing the NEMS-GI and REMI models individually and discusses how the two models have been integrated into NEMS-REMI.  The scenario under consideration here, the imposition of a carbon tax on greenhouse gas emissions, is then explained.  Next, the macroeconomic impact results in terms of change in output and employment are presented graphically and the sources of the macroeconomic variation across U.S. regions are discussed.  Finally, the conclusion shows how this comprehensive tool will be useful to answer “what if” questions regarding the macroeconomic impacts of energy and environmental legislation by industry and region for public and private decision-makers.

II.   Integration of NEMS-GI and REMI PI+ Models

NEMS-GI Energy Model

The National Energy Modeling System (NEMS) is a computer-based, structured modular system that models the energy-economy of the US to 2030.[i] Originally designed by the Energy Information Administration (EIA), the NEMS model has been modified by SAIC to enhance its applicability in forecasting in U.S. energy markets and policy studies.  With an Integrating Module and a series of relatively independent modules that represent the domestic energy system, the international energy market, and the economy, the NEMS-GI model projects the production, demand, consumption, imports, and prices of energy into the future.  Given that energy markets are heterogeneous, the NEMS-GI model uses a modular design that allows for the methodology appropriate to the supply, conversion, and end-use demand sectors in a given market of a given region.[ii]

The NEMS-GI system consists of five modules with multiple sub-modules.[iii] First, the Demand module has four modules of end-use energy demand (Residential Demand, Commercial Demand, Industrial Demand, and Transportation Demand) and uses the nine U.S. Census divisions.  Secondly, there are two Conversion modules: the Electricity Market and Petroleum Market.  The former module uses 15 supply regions based on the North American Electric Reliability Council (NERC), while the latter uses five regions based on the Petroleum Administration for Defense Districts.  Thirdly, there are four Supply modules: Oil and Gas, Natural Gas Transmission and Distribution, Coal Market, and Renewable Fuels. The Oil and Gas Supply Modules use 12 supply regions, including three offshore and three Alaskan regions, while the Natural Gas Transmission and Distribution uses 30 regions, including 10 pipeline border points and eight LNG import regions.  Fourth, there is one International Energy Module to simulate international energy markets.  Fifth, there is the Macroeconomic Activity Module that simulates energy/economy interactions.

REMI PI+ Macroeconomic Model

The REMI Policy Insight is a dynamic, structural economic modeling, forecasting and policy analysis model that contains detailed industry sectors and provides regional as well as national-level analysis and projections.  The PI+ model integrates input-output, computable general equilibrium, econometric and economic geography methodologies.  REMI simultaneous equations represent fundamental relationships between key economic and social variables and yield response estimates.

The REMI model contains a five-block structure that represents the entire macro-economy, as shown in Figures 1 and 2.[iv] Block 1, Output and Demand, includes output, demand, consumption, investment, government spending, commodity access, import and export concepts.  Block 2, Labor and Capital Demand, includes labor productivity, labor intensity, and the optimal capital stocks.  Block 3, Population and Supply, includes detailed demographic information about a given region.  Block 4, Compensation, Prices and Costs, includes delivered prices, production costs, equipment cost, the consumption deflator, consumer prices, the price of housing, and the compensation equation.  Economic geography concepts account for the productivity and price effects of access to specialized labor, goods, and services.  In Block 5, Market Shares equations measure the proportion of local and export markets that are captured by each industry.

The dynamic REMI PI+ model also captures the spatial dimension of the economy by its economic geography equations that incorporate transportation costs, industry clusters, and economic activity agglomeration effects.[v] PI+ multi-regional models also have interactions among regions such as trade and commuting flows.[vi]

For the integration with the NEMS energy model, a multi-region national version of the PI+ model is used allowing for macroeconomic analysis on both national and the 9 census-regions levels.  The private nonfarm sector of the economy is divided into 165 private industry sub-sectors based on the North American Classification System (NAICS).

Figure 1. REMI Model Linkages


Figure 2. PI+ Economic Geography Linkages

Integration of the Models

In order to demonstrate the enhanced results that the integration of the two models yields, we project the economy-wide effects of changes in GDP and employment in U.S regions due to a carbon tax.  A carbon tax represents a potential policy action to discourage and reduce the generation of greenhouse gas emissions.[vii] This is to be achieved by charging carbon dioxide emitters a tax per ton of CO2 gas emissions they produce and therefore increase the price of carbon-based energy.  The tax revenue generated by such a policy may then be used by the government for deficit reduction, increased government spending, tax reductions or transfers.

The NEMS-GI and REMI PI+ models are integrated as follows: First, a policy change represented by a tax imposed per ton of greenhouse gas emission increasing the Carbon Allowance Price, is run through the NEMS-GI model.  This policy results in higher fuel prices across the nine census regions.  Then, these regional fuel-specific price changes are entered into the REMI economic impact model in order to identify the iterative cumulative effects on the U.S. economy.

The integrated NEMS-REMI model follows a two-step process to estimate how the imposition of a carbon tax on greenhouse emissions would affect the US economy and the industries on a regional basis.

Step 1 calculates the three primary direct effects of this policy initiative in the NEMS model:

  • Changes in Energy Demand by Industry: A major policy change such as a carbon tax will lead to significant change in demand for the products of five energy industries across the country in all regions: Crude Petroleum, Coal Mining, Electric Utilities, Gas Utility, and Petroleum Refining.
  • Changes in Energy Prices: A carbon tax imposed on greenhouse gas emissions will add to the business costs for three types of fuel for industrial and commercial users: Electricity, Natural Gas, and Residual Fuels.  It will also add to the household costs for four types of fuel consumption categories: Gasoline and Oil, Fuel Oil and Coal, Electricity, and Gas, resulting in higher energy prices across the U.S.
  • Transfer Payments: A carbon tax imposed on per ton of gas emission will result in significant government tax revenue.  This scenario assumes that the increased revenue is distributed to U.S. residents as transfer payments.

Step 2 takes these three primary effects of the policy change as calculated by the NEMS-GI model, i.e., the NEMS output, and inputs them into the REMI PI+ model.  Using the changes in energy demand, energy prices, and transfer payments, this dynamic, general equilibrium, macroeconomic model of the US economy calculates the changes in GDP and employment both nationally and on a regional basis for the nine census regions. Results are also obtained for the change in employment and output by industry over time through 2030.

Figure 3 provides a schematic overview of the integration of the NEMS and REMI models and shows how the final GDP, output and employment results, among others, are obtained given the carbon tax scenario.

Figure 3.  Model Structure

III.   Integration Results

The implementation of a national carbon tax generates energy price and energy consumption changes in the NEMS-GI model.  The data on energy consumption by sector and source and energy prices by sector and source are used to calculate the increase in energy prices for each of the nine Census Divisions (Figure 4).  In addition, the demand for energy producing sectors is reduced to reflect the change in energy consumption.

Figure 4.  Model treatment of impact on energy prices

  • Since this scenario uses the energy consumption change generated by NEMS-GI, REMI’s fuel substitution module is implemented in order to turn off the endogenous response to changes in fuel prices.
  • The Instantaneous Product Market Clearing (IPMC) alternative model option is used based on the assumption that businesses and consumers will have knowledge of the tax prior to being subject to it, and therefore will have time to adjust behavior (energy consumption).  The IPMC option uses the current year actual values of production costs instead of the moving average (which is lagged).
  • The Labor Intensity Response to relative factor costs is disabled to more closely approximate the macroeconomic response of the NEMS-GI model.
  • The Investment Stock Adjustment Process is disabled and replaced with an Investment Response to the level of Activity to adjust the timing of investment to more closely approximate the macroeconomic response of the NEMS-GI model.

In this scenario, the carbon tax revenue generated by a tax per ton of carbon emission is estimated in the NEMS-GI model.  It is assumed that the entire tax collected is then returned to US resident as transfers (“dividends”), which are distributed on a per-capita basis.

The carbon allowance price is projected to increase gradually over time from $70 per metric ton CO2 in 2012 to $170 in 2030 (Figure 5).  Thus, the carbon tax revenue generated by this policy initiative rises over time, leading to greater transfers returned to US residents, i.e., consumers. As a result, consumption is projected to increase nationally.

Figure 5. Carbon Tax Per Ton of Greenhouse Gas Emissions Over Time

Source: SAIC

National Results

The implementation of a carbon tax on greenhouse gas emissions initially causes a decline in National GDP and employment at an increasing rate until 2017-2018, followed by declines at a decreasing rate until they begin to grow at a positive rate close to 2030.

The carbon tax would initially cause a change in the National GDP of -0.75% according to the REMI model and of -1% in the NEMS-GI model.  The decline in GDP continues to increase until around 2017, when the GDP growth rate falls to -2.25% in REMI PI+ and -2.50% in NEMS.  After this point of inflection, GDP begins to decline at a decreasing rate.  The GDP growth rate returns to positive territory around 2028 in the NEMS-GI model, while in the REMI model this rate continues to be negative at slightly less than -1% in 2030.

The change in national employment as a result of an increasing carbon tax follows a parallel path in the two models.  Employment initially declines annually by approximately 0.40% and then continues to decline at an increasing rate until around 2016-2017 when the growth rate reaches its trough at -1.60% in the REMI model and -1.80% in NEMS.  After this inflection point, employment declines at a decreasing rate until 2030 when the annual change is projected to be -0.20% in the REMI PI+, while it is slightly higher at -0.30% in the NEMS-GI model.

Implementing a carbon tax would affect industry users of energy nationally by raising the production cost of energy and reduces output, which is ultimately passed onto consumers in the form of higher prices for goods and services.  As the price of energy increases, demand would decline as both private consumers and businesses modify their behavior to cut their costs.  The net positive change in energy prices will induce consumers of all classes to reduce consumption of other goods as well.  As companies produce less and consumers spend less, fewer workers are hired, resulting in lower employment.  However, as tax revenue is distributed to consumers in the form of dividends, consumption is likely to recover over time to its previous levels of zero carbon tax, with concomitant increases in GDP and employment.

Industry-Level Results

On an industry-level, the carbon tax is projected to cause a shift in industry composition in the U.S. from energy-intensive industries such as Manufacturing as well as the obvious energy-producers such as Oil & Gas and Utilities to service sectors such as Retailing.  Thus, the tax is expected to accelerate the ongoing trend in the sectoral composition of the U.S GDP and employment in the form of increasing share of the service sector and decline in the share of manufacturing.

Throughout the country, the carbon tax will increase energy prices, albeit to varying degrees in different regions.  Higher energy prices are projected to make US industries less competitive internationally.  Thus, the value of US exports will also decline over time, as adjustments required for achieving competitiveness in production under higher energy costs will take time.

Regional Results

The macroeconomic impacts of implementing an increasing carbon tax follow a similar trend over time nationally and across the nine census regions.  The initial drop in GDP and employment caused by the initial tax shock to industry production due to the increased cost of production, higher energy prices, and higher taxes is followed by a gradual recovery of declining negative growth and then increasing rates of positive growth.  However, there is significant variation across regions as to how well they perform over time and cope with the initial drop in demand and increase in prices of energy and energy-intensive goods.

Though following the same trend over time of initial hit and gradual recovery, New England performs best among all regions under the carbon tax scenario in the REMI model.  The magnitude of the initial drop in the Gross Domestic Product in New England in the couple of years following the enactment of the tax does not exceed -1.0 %.  Then, the negative trend in the percentage change in GDP is reversed around 2017.  The decline in GDP growth gets less and less, until it inches into the positive territory around 2027, and then keeps rising.  Similarly, employment experiences an initial decline in New England, which slightly exceeds the -0.50% mark at its trough.  Around 2015, it ceases to decrease at an increasing rate.  The drop in employment gets less and less, until employment starts increasing again around 2019.  Continuing on a positive growth trend, employment is projected to grow by almost 1% by 2030.  This trend may be explained by the composition of industry in New England: more specifically, there are not as many energy-producing or energy-intensive industries in the region.  In addition, the gain in service sectors such as retail is enough to offset the losses in energy-intensive industries.

The projected trend is similar for energy-intensive West South Central, which contains oil-producing and refining Texas and Louisiana, but the magnitude of the drops in Gross Domestic Product and employment in the region is decidedly more negative for this region in the REMI model.  Furthermore, although both GDP and employment stop declining at an increasing rate after a period of five years, they still continue to experience negative growth until 2030 and possibly beyond.  The initial decline in GDP in West South Central exceeds -4% from 2013 through 2020.  Although GDP stops contracting at an increasing rate after 2017, the GDP growth rate is still slightly less than -3.5% in 2030.  The trend of change in employment over time in the region is slightly less negative, but it follows a similar path.  There is sharp decline in employment initially in West South Central, with the growth rate plunging from -1.5% to almost -3%.  The growth rate continues to decrease until around 2017, when it is close to -3.25%.  The increasing rate of the decline is then reversed and the trend line inches upwards.  However, the growth rate of employment remains in the negative territory.  For example, the percent change in employment in 2030 is slightly less than -2%.

The West South Central region experiences the most significant shock in terms of the drop in GDP and employment in the REMI model because it is heavily energy-producing and energy-intensive in terms of industry composition.  These sectors are the ones that experience higher business costs and less demand for their products as a result of an increasing carbon tax imposed on CO2 emissions and therefore, they take the greatest hit by the tax.  Energy-producing regions such as West South Central are likely to experience the greatest decline in GDP and employment.

A comparison of the regional results reveals that although GDP and employment follow the same trajectory in the two models, the REMI and NEMS models yield results of differing magnitude.  This may be clearly observed in comparing the results for the New England and West South Central regions.  Although New England performs the best in terms of weathering the initial negative shock and later economic recovery among all regions in the REMI model, West South Central actually fares better than New England in the NEMS-GI model.  In the REMI model, however, the West South Central Region sees a drop of 4% in employment and never fully recovers.  On the other hand, New England experiences a much lower decline and ultimately recovers back into a positive growth rate.  Given that West South Central is both significantly a more energy-producing and energy-intensive area than New England; that the latter would be in a better shape to adjust to a gradually increasing carbon tax makes economic sense.

In fact, taken as a whole, one does not observe differences of a significant magnitude among the trajectories of the 9 regions in the NEMS-GI model.  The changes in GDP and employment move in sync for all of them and the range of differences does not exceed 1%.  On the other hand, in the REMI model, the economic changes are projected to vary in the range of 2 to 3% among the regions.

The differences in the projected macroeconomic effects between the two models may be attributed to the difference in their fundamental approaches.  The NEMS-GI model essentially has a top-down approach to modeling change across the nation.  More specifically, the policy changes are implemented in the NEMS-GI model, yielding a national outcome.  The average national effects are then shared out among the regions according to their share in population, yielding economic changes moving in sync across all regions.  The REMI model, on the other hand, takes a bottom-up approach to modeling regional economic changes.  Thus, economic changes are modeled in the specific regions where they occur in the REMI PI+ model.  Therefore, region-specific results are obtained to chart the trajectory of GDP, employment, population, and prices among other variables.  Furthermore, these results factor the interaction effect among regions into account.  As a result, the REMI PI+ model is able to yield more accurate, longitudinal trajectories of macroeconomic variables for each U.S. region caused by a major policy implementation such as a carbon tax.

Table 1. NEMS-REMI Regional Projections

  1. REMI Percentage Change in Employment, Regional

% Change in Employment, New England

% Change in Employment, West South Central

  1. REMI Percentage Change in Output, Regional

% Change in Output, New England

% Change in Output, West South Central

 

Figure 6. NEMS-GI and NEMS-REMI Projections of Change in Real National GDP


Figure 7. NEMS-GI and NEMS-REMI Projection of Change in National Employment


Figure 8. NEMS-REMI Projection of Percent Change in Real GDP for US Regions


Figure 9. NEMS-REMI Projection of Percent Change in Employment for US Regions


Figure 10. Projected Change in Employment for New England


Figure 11. Projected Change in Employment for West South Central

IV.   Top-Down vs. Bottom-Up Approach to Energy/Economic Modeling

Energy/economic modeling approaches can be described as top-down, bottom-up, or hybrid approach.  Top-down modeling use aggregate energy and economic data, typically at the national level, to forecast changes in energy markets and the economy in response to policy actions.  Bottom-up modeling represents differing technologies and economic structures at a disaggregate level; aggregate results are determined as the sum of disaggregate changes.

The NEMS-GI energy model is constructed as a bottom-up model, with energy markets represented at the spatially disaggregate level.  While region-specific modeling of energy markets requires extensive data collection and model development, NEMS-GI has been developed along these lines due to the inherent regional nature of energy markets.  For example, electric generation technologies and fuel types vary dramatically from one power generation region to another; additionally, the availability of renewable and non-renewable resources such as wind and coal vary significantly on a spatial level.

As in energy markets, economic structures vary greatly on the spatial level.  Energy-producing and energy-dependent industries are heavily concentrated in states such as Texas, Louisiana, West Virginia, and Wyoming.  Thus, changes in energy policies are likely to impact these states differently than less energy-intensive states such as New York and California.  Due to the interactions between industries and other parts of the economy, direct changes to energy prices and outputs result in greater indirect and induced effects on the state or regional economy.  Additionally, regional economies are open with respect to trade and migration; thus, macroeconomic impacts for a given region will affect other regions through trade flows and movement of labor.

The macroeconomic modeling aspect of NEMS-GI uses a top-down approach.  Direct price, investment, and other changes that occur in energy markets are aggregated into national macroeconomic model inputs, which then simulate national-level macroeconomic changes.  These national macroeconomic impacts are proportionally shared out to the census-region level to estimate regional economic effects.

With NEMS-REMI, regional changes in energy markets, the output of NEMS, are used as economic policy variable inputs at the regional level.  Thus, the spatial detail of NEMS output matches the spatial configuration of REMI PI+.  Changes in prices, investments, and fuel demand that occur at the census region drive changes in the regional economy.  Since policy variable inputs differ markedly from one region to another, the spatial match is important in capturing their effects on the regional economic level.

V. Conclusion

This paper has shown how the integration of the NEMS-GI Energy Model with the REMI PI+ Macroeconomic Modeling System provides a comprehensive tool that measures the macroeconomic impacts of policy initiatives in energy and environmental legislation.  By inputting into the PI+ model the NEMS-GI model output of changes in energy prices and demand across the U.S caused by a hypothetical carbon tax that increases the carbon allowance price per ton of greenhouse gas emission, users are able to estimate and analyze the impact of a change in the energy market on the GDP and employment on a regional level by industry in the NEMS-REMI model.

The strength of the REMI PI+ model is the ability to conduct a macroeconomic analysis on a regional as well as national basis that is suited to the industry composition and other economic characteristics of a given region and to provide region-specific findings on a detailed industry level.  PI+’s bottom up methodology allows it to analyze the macroeconomic effect of changes in policy in the specific regions that they occur and also fully factor in the interaction effect between regions.  Hence, the NEMS-REMI model yields the most accurate results for each region on the effects of policy changes in energy markets.

Energy and Environment constitute two of the most important policy areas of the agenda on Capitol Hill.  Policymakers are looking at various ways to reduce the carbon emissions of the nations with the least adverse economic impact possible.  By successfully conjoining the NEMS-GI model with the REMI model, public and private decision-makers will have a much better understanding of the wide-ranging economic effects of proposed energy and environmental policy initiatives, especially in terms of their impact on output and employment, both regionally and nationally.  Using this integrated tool, individuals and institutions can more effectively project, shape, and respond to changes affecting energy costs and environmental impacts.  Given the importance of this issue to economic and strategic security, it will be vital going forward that such sound, detailed analyses drive decision-making at the highest levels that affect U.S. regions and the country as a whole.

 

REFERENCES

Energy Information Administration, October 2009, The National Energy Modeling System: An Overview 2009. http://www.eia.doe.gov/oiaf/aeo/overview/

Regional Economic Models Inc, 2009, REMI PI+ V 1.1Model Equations. http://www.remi.com/uploads/File/Documentation/PI+_v1-1_Model_Equations.pdf

George I. Treyz, 1993, Regional Economic Modeling: A Systematic Approach to Economic Forecasting and Policy Analysis, Boston: Kluwer Academic Publishers.

Ilya Leybovich, March 17, 2009, Carbon Tax vs. Cap and Trade, Industry Market Trends. http://news.thomasnet.com/IMT/archives/2009/03/carbon-tax-versus-cap-and-trade-system-debate-heats-up.html

Janet E. Milne, September 5, 2008, The Case for Carbon Taxes: Simple is Better, Bulletin of the Atomic Scientists.

 

* This article copyrighted and reprinted by permission from the International Association for Energy Economics. The material first appeared in the online proceedings of the 33rd IAEE International Conference, Rio de Janeiro, June, 2010.

** Frederick Treyz is Chief Executive Officer of Regional Economic Models, Inc. (REMI), 433 West St Amherst, MA 01002 USA, E-Mail: fred@remi.com, Phone: +1-413-549-1169 x 233.  Sherri Lawrence is Senior VP and Sinan Hastorun is Associate Research Economist at REMI.


[i] The National Energy Modeling System: An Overview 2009, Energy Information Administration, October 2009, p. 1.

[ii] The National Energy Modeling System, p. 3.

[iii] The National Energy Modeling System, p. 6.

[iv] REMI PI+ V 1.1 Model Equations, Regional Economic Models Inc, 2009, p. 3.

[v] Regional Economic Modeling: A Systematic Approach to Economic Forecasting and Policy Analysis, George I. Treyz, 1993, Boston: Kluwer Academic Publishers.

[vi] REMI PI+ V 1.1 Model Equations, p. 4.

[vii] Carbon Tax vs. Cap and Trade, Ilya Leybovich, Industry Market Trends, March 17, 2009.

Click to view a printable version of this article.