1 Designing Climate Change Scenarios in a Global Economic Model Warwick J McKibbin ANU, Lowy Institute and Brookings Prepared for the OECD conference on.

1

Designing Climate Change Scenarios in a Global Economic

Model

Warwick J McKibbinANU, Lowy Institute and Brookings

Prepared for the OECD conference on “Global Convergence Scenarios: Structural and Policy Issues” to be held in Paris, January 16, 2006

2

• Based on 2 papers:

“Long Run Projections for Climate Change Scenarios”

McKibbin, Pearce and Stegman (2004)

“Convergence and Per Capita Carbon Emissions”

McKibbin and Stegman (2005)

3

Structure of Presentation

• Overview– Why Emission projections matter– Are Projections Useful?

• What do We Know about projecting the future?– Looking for Empirical regularities

• Some Theoretical Issues– Sources of growth

• Convergence (of what?) Across countries• A common approach used in energy models• The G-Cubed Economic Approach of making projections

– Sensitivity to PPP versus MER convergence assumptions• Is there a Better Way to make projections for climate policy?• Conclusion

4

Why emission projections matter

• Critical input into climate change debate– Policies have been and are being conditioned on the

baseline and initial conditions• Emission projections feed into climate models to make

temperature projections• Temperature projections feed into impact models to

assess – environmental/ecological/economic/health impacts over the next century

5

Are Projections Useful?

• Yes but– They are but we shouldn’t believe too much outside of

the next 30 years or so– They should be based on the best empirical evidence

and best practice– They should reflect the underlying uncertainty that is

manifest in projecting the future

6

What do We Know?

• We have about 60 years of data to look for patterns in the data and test hypotheses

– Economy wide responses to changes in energy prices– Determinants of growth– Patterns of convergence

7

What do we Know?

Oil price shocks of the 1970s generated important information for estimating the impacts of energy prices on economic behavior

- Supply (substitution, technical change)- Demand (conservation, substitution)

Figure 1: GDP, Energy and Emissions for US and JapanIndex Numbers, 1965=1

0

1

2

3

4

5

1965 1970 1975 1980 1985 1990

US GDP Japan GDP

US Energy Japan Energy

US Emissions Japan Emissions

9

Interpretation

• Economic Modelers use this as evidence that relative prices matter – (and estimate the effects)

• Energy modelers tend to use the data post 1975 to calculate “Autonomous Energy Efficiency Improvements”

• In projecting the future, it matters a great deal which approach is followed.

10

Theoretical Issues in Forecasting Growth

• Sources of output growth within a country– Increases in the supply capital, labor, energy,

materials– Increase in the quality of these inputs– Improvements in the way the inputs are used

(technical change)– Improvements in the way inputs are allocated across

the economy– Improvements in the way inputs are allocated across

the world

11

Theoretical Issues in Forecasting Global Growth

• Convergence across countries– What converges?

• Incomes per capita• GDP per capita• Aggregate level or rate of technical progress• Sectoral level or rates of technical progress

– The empirical literature examines conditional versus unconditional convergence of income per capita and to a lesser extent output per worker (productivity)

– Little empirical evidence of unconditional convergence across large numbers of countries

12

Approaches

• Many energy models use assumption about emissions per capita converging or energy efficiency converging autonomously and then overlay this with aggregate GDP projections

• Hence the reason why the assumptions about economic growth and the PPP debate don’t matter much in these models. You just change the numeraire.

13

Do Carbon Emissions per Capita converge?

Some models assume this either as fact or as a desired target

0.0

0.5

1.0

1.5

2.0

1950 1955 1960 1965 1970 1975 1980 1985 1990 1995

STDEV

VAR

CV

MEAN

Figure 1: Summary Measures of Spread Emissions Per Capita

15

• Need to be careful how data is interpreted

• McKibbin and Stegman (2004) use dynamic kernel estimation to explore convergence

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0 1 2 3 4 5 6

1950 1960

1970 1980

1990 1999

Figure 3: The Cross-Sectional Distribution of Emissions per Capita Sample A

Metric Tons of Carbon Per Capita

f

17

Does GDP per capita converge?

Some models assume this either as fact or as a desired target

0.0

0.2

0.4

0.6

0.8

1.0

0 1 2 3 4 5 6

1971 1980 1990 2000

Relative GDP Per Capita

Figure 10: The Cross Country Distribution of GDP Per Capita Density Estimates

f

19

The G-Cubed Approach

McKibbin & Wilcoxen

20

The G-Cubed Model

– Countries• United States• Japan• Australia• New Zealand• Canada• Rest of OECD• Brazil• Rest of Latin America• China• India• Eastern Europe and Former Soviet Union• Oil Exporting Developing Countries• Other non Oil Exporting Developing Countries

21

The G-Cubed Model

– Sectors– (1) Electric Utilities– (2) Gas Utilities– (3) Petroleum Refining– (4) Coal Mining– (5) Crude Oil and Gas Extraction– (6) Other Mining– (7) Agriculture, Fishing and Hunting– (8) Forestry and Wood Products– (9) Durable Manufacturing– (10) Non Durable Manufacturing– (11) Transportation– (12) Services– (Y) capital good producing sector

22

Features of the G-Cubed Model

• Dynamic • Intertemporal • General Equilibrium • Multi-Country • Multi-sectoral • Econometric • Macroeconomic

23

G-Cubed Approach of Generating Future Projections

• Make assumptions about labor augmenting technical change (LATC) for each sector in the US

• Calculate economy wide gaps between LATC within each sector relative to the US sector such that the TFP gap across sectors is approximately equal to the PPP GDP per worker gap

• Assume that the gap in LATC between each country and the US closes by x% per year (we vary this between 0 and 2%)

24

Process of Generating Future Projections

• Assume that labor supply grows at the rate of the mid range UN population projections from 2002 to 2050 and then gradually converges across countries to zero population growth in the long run.

• Other exogenous inputs include tax rates per country per sector, tariff rates per country per sector, monetary and fiscal regimes

25

Process of Generating Future Projections

• Given initial capital stocks in each sector, the overall output growth rate of an economy depends;– the growth on LATC (exogenous), – labor force (exogenous in the long run); – the accumulation of capital (endogenous)– the use of materials input by type (endogenous)– the use of energy inputs by type (endogenous)

26

Key Points

• The projection of carbon emissions will depend on the growth of the demand for carbon intensive inputs (oil, natural gas, coal).

• There is no reason for a fixed relationship between growth in GDP and growth in carbon emissions

• The is no need for carbon emissions per capita to converge unconditionally

• The outcomes depend on the trend inputs and the structural change in the economy induced on the supply side and demand side of all economies.

27

Key Points

• For global emissions it matters which sector in which country experiences productivity growth

• Models that assume constant ratios or linear trends between GDP and emissions are likely to be problematic if the actual growth process involves structural change

28

Theoretical Issues

• PPP versus market exchange rates– Castles and Henderson argue that if the rate of

growth of developing countries are measured based on the initial differences in income per capita then it is critical to measure this gap using PPP

– Many studies use market exchange rates and so growth is likely to be overestimated.

– Does this matter?• An empirical question

29

PPP versus Market Exchange Rates

• G-Cubed uses a PPP concept for GDP to benchmark the initial productivity gap between sectors in each country relative to the US

• The rate of economic growth and emissions outcomes are then determined simultaneously by the model

• Suppose we use market exchange rates to benchmark initial gaps between countries – what difference does this make to emission projections?

30

How much does PPP versus MER Matter?

• The ratio of the productivity of China to the US is 0.2 based on PPP

• The ratio of the productivity of LDCs to the US is 0.4 based on PPP

• Suppose we assume – China has an initial gap of 0.1 (from MER)– LDCs have a gap of 0.13 (from MER)

Change in Carbon Emissions Market vs PPP2050 and 2100

0

20

40

60

80

100

120

US Japan Australia Europe ROECD China LDCs EEB OPEC World

2050 2100

32

Implications

• PPP versus market exchange rates makes a big difference to the projections of economic growth and the projections of future carbon emissions

• The errors affect both developing and developed countries

• Does this matter for temperature?– Manne and Richels argue that temperature is based

on the stock of cumulative emissions and flows take time to have any impact

– BUT these magnitudes are too large for the IPCC to dismiss the way they have to date.

33

The Response of the SRES Authors to Critiques

• Convergence is not assumed in most scenarios (not clear what is assumed)

• Doesn’t matter whether convergence is defined in PPP or MER one can always convert between the two (problem is that there is no empirical relationship)

• Even where convergence is assumed, it is not clear that assuming high growth in developing countries would cause emissions to be overestimated because with higher income there would be more investment in technology and more emissions reductions– How plausible is this argument?

34

Is there a better way to undertake projections of the world economy for climate evaluation?

• Focus on the time frames we understand better and separate clearly the types of uncertainty

– The past | the near future | the distant future

• Climate models can give some indication of what the actual emissions in the past until now would do to the climate in future years

• Economic models based on the past 60 years of data allow us to project about 20-30 years into the future with some empirical basis (rather than pure speculation) and allows statistical uncertainty to be computed– This gives us another 30 years of data to add to the climate

projections with a different degree of confidence.

35

Finally

• Projecting the world economy over the time horizons required for making temperature projections is not easy

• It is a big mistake to rely on the accuracy of these projections in formulating and conditioning policy

• A better approach in the McKibbin-Wilcoxen Blueprint which focuses on costs and benefits rather that target and timetables

36

Background Papers

www.gcubed.com

www.sensiblepolicy.com