Investments in Global Warming Mitigation: The Case of “Activities Implemented Jointly”

Policy Sci (2006) 39:233–248

DOI 10.1007/s11077-006-9020-9

ORIGINAL ART ICLE

Investments in global warming mitigation: The case of“activities implemented jointly”

Nives Dolsak · Maureen Dunn

Received: 17 November 2005 / Accepted: 17 July 2006C© Springer Science + Business Media B.V. 2006

Abstract This paper examines bilateral cooperation between developed countries (home

country) and developing countries (host country) to reduce greenhouse gas emissions and

to enhance carbon dioxide sinks. With the home-host country pair as the unit of analysis,

our logistic regression model examines 158 Activities Implemented Jointly (AIJ) investment

projects from 1993 until 2002 across 2541 country-pairs. Because the marginal costs of

reducing emissions may be lower in developing countries, the AIJ projects served as a policy

laboratory to assess whether such investments might be advantageous to both countries in

the event future regimes allowed emission credits from such bilateral projects. Instead of

investing in home countries where maximum pollution reductions (or carbon sequestration)

might be possible, home countries invest in locations where they can conduct their policy

experiments at low transaction costs. Prior trade and aid relationships were used as a proxy.

Regarding energy projects, location decisions are driven by home countries’ desire to reduce

air pollution that they receive from abroad. Geography – proximity of a host country to

a home country – in interaction with host country’s coal production, is a very important

driver of location decision in AIJ energy sector projects. Location of sequestration projects

is impacted by the host country’s potential for avoiding deforestation as well as by previous

aid and trade patterns between a home and a host country. Proximity is not important in this

case.

Keywords Global climate change policy . Global warming policy . International

environmental policy . International environmental regimes . International cooperation .

Global commons . Energy policy . Environmental policy . Activities implemented jointly .

International regimes

Introduction

Due to the open access attribute of the global atmosphere, unilateral action by any country

is unlikely to lead to significant reductions in greenhouse gas emissions and multilateral

N. Dolsak · M. DunnUniversity of Washington-Bothell

Springer

234 Policy Sci (2006) 39:233–248

action is likely to become the preferred mode of action. In 1992, 150 countries negotiated

and agreed to an international environmental regime aimed at stabilizing emissions of green-

house gases.1 This regime was embodied in the United Nations Framework Convention on

Climate Change or the UNFCCC. The UNFCCC sought to first stabilize and then reduce

global greenhouse gas emissions. Developed countries accepted their historical responsibility

for global climate change and committed to stabilizing their emissions in the future while de-

veloping countries were not expected to adopt costly emission reduction policies until much

later.

The UNFCCC devised country-specific emission reduction goals and outlined mecha-

nisms through which to achieve these goals. Bilaterally negotiated projects – the Activities

Implemented Jointly or AIJ projects – where a home country invests in a host country were

the key mechanisms for emission reductions as well as carbon sink enhancement. AIJ projects

preceded other flexibility mechanisms defined subsequently in the Kyoto Protocol, including

Clean Development Mechanisms (CDMs) and Joint Implementation. In some ways, AIJ can

be seen as a “laboratory” for these subsequent mechanisms (Michaelowa, 2002). While AIJ

projects did not allow any emission credits to be issued, it is likely that home countries saw

them as providing information relevant for future CDM projects, even legitimizing these

projects as a viable option for the implementation of the Kyoto protocol. As reported by

Michaelowa (2002), in the case of the Netherlands’ AIJ project in Poland, the agreement on

sharing emission reduction credits was already negotiated at the AIJ stage. Nevertheless, the

absence of emission credits eventually made AIJs less attractive to developed countries and

led to their replacement by CDMs that provided such credits.

How did home countries identify host countries for AIJ projects? Drawing on the inter-

national aid and international regime literatures, this paper investigates the location puzzle

by examining characteristics of home-host country pairs and characteristics specific to coun-

tries involved in the AIJs and comparing them to country-pairs with no projects. Because

marginal costs – economic and political – of reducing emissions may be lower in developing

countries, the AIJ projects served as a laboratory to examine whether such investments might

be advantageous to both countries, in the event future regimes granted emission reduction

credits to similar bilateral projects. Simply put, it seemed worthwhile to test mechanisms

through which developing countries could gain scarce capital and energy efficient technolo-

gies while developed countries would meet reduction targets at a lower cost. Indeed, the Clean

Development Mechanism proposed under the 1997 Kyoto protocol is based on this logic; it

seeks to create a win-win situation for the host and the home countries and at the same time,

lead to the production of an open access global public good: curbing global warming. Thus,

AIJ projects served as the laboratory for country pairs to explore the benefits and costs of

common endeavors in the future.

While AIJ policy experiments might appeal to several actors, it is not clear as to what

factors might influence home countries’ location decisions. Would factors not related to

global warming also influence location decisions? For example, home countries might view

AIJs as foreign aid and identify AIJ locations based on philanthropic or strategic reasons.

Or perhaps, AIJ projects might be located with an objective to corner tangible benefits for

the home country – such as, reduction of pollution at home – rather than aim to maximize

reduction in green house gas emissions. This might make the policy experiment worthwhile

on its own accord even if emission reduction credits were not provided by new regimes in

the future.

1 Greenhouse gases that are included in the regime are carbon dioxide, methane, and chlorofluorocarbons.

Springer

Policy Sci (2006) 39:233–248 235

Indeed, prior to the Kyoto Protocol, a number of home countries invested in bilateral

projects abroad with the stated objective of reducing emissions of greenhouse gases and/or

enhancing carbon dioxide sinks. These AIJ projects began in the mid 1990s. More than half

of the 158 AIJ projects examined in this study are in the energy sector and the rest in forestry

(carbon sequestration), agriculture (carbon dioxide and methane emission reductions) and

landfill management (methane emission reductions). These projects vary in terms of their

activities, technology, projected reductions in carbon dioxide emissions, project lifetime, and

costs. Here are some examples:� In the Jelgava Energy Efficiency project between Sweden and Latvia, the Swedish National

Board for Industrial and Technical Development financed an energy project for a school in

town Jelgava. This project included renovation and insulation of the roof and the installation

of a heat exchanger technology (UNFCCC, 2002a). The project started in 1995 and was

expected to finish in 2000 costing $150,000. The estimated life-time reduction of emissions

of carbon dioxide equivalent is 400 metric tons.� In the Model Project for Energy Conservation in Electric Furnace used for Ferro-AlloyRefining, Japan invested in improving energy efficiency in metal industry in China. The

stated goal of the project was: “to contribute to efficient use of energy and consequently

protection of the local environment in People’s Republic of China as well as the reduc-

tion of CO2 emission. . .and disseminating the technology in People’s Republic of China”

(UNFCCC, 2002b: 1). The total estimated reduction of carbon dioxide equivalent over the

life-time of the project is about 290,500 metric tons (UNFCCC, 2002b).� In fuel switching project in the city of Decin, Czech Republic, a number of investors from

the U.S, including Wisconsin Electric Power Company, Commonwealth Edison Company,

and NIPSCO Development Company co-financed a new district heating plant. The total

estimated reduction of carbon dioxide equivalent over the life-time of the project is about

607,000 metric tons (UNFCCC, 2002c).� In Plantas Eolicas S.A. Wind Facility, Costa Rica, Northeast Utilities (a U.S. company)

and Charter Oak Energy (also associated with Northeast Utilities) built a 20 megawatt

wind facility. According to the UNFCCC information on AIJ projects (UNFCCC, 2005),

this facility has been in operation since 1996. The wind electricity generation facility is

displacing fossil-fuel electricity generation thereby reducing greenhouse gas emissions.� In the Klinki Forestry Project, a carbon sequestration project negotiated between the gov-

ernments of the U.S. and Costa Rica in 1995, marginal agricultural lands in Costa Rica

are converted to commercial tree plantations (UNFCCC, 2005; Reforest the Tropics, Inc.,

2006). Reforest the Tropics, a non-profit organization, was selected to manage the project.

The host of the project is usually a developing country or an economy in transition. The

home country is always a developed country (Annex I country in the UNFCCC language),

but a variety of home country actors can be involved. We classify them broadly as private

sector actors and government/NGO actors:� Some home countries designated a special governmental office to engage in AIJ projects,

for example Sweden and Switzerland. All country-pairs with AIJ projects with these home

countries were coded as government.� Some countries used existing environmental or international affairs offices. In this case,

the country-pair was coded as government.� Some countries experimented with both governmental and private sector driven AIJ projects

(for example, Belgium, France, the Netherlands, and the USA). Here, depending on the

type of the actor, the country-pair was coded correspondingly. In some instances both

Springer

236 Policy Sci (2006) 39:233–248

Table 1 Number ofcountry-pairs with an AIJ projectby the type of project and thehome country participant

Government and NGO Private sector

Energy 43 27

Sequestration 9 12

government and private actors were involved in projects in a given country-pair. For ex-

ample, both U.S. government and U.S. industry were involved in projects in Costa Rica.

In such cases, the country-pair was assigned both codes.� In yet another group of countries, energy industry in home countries was the key actor.

These include Canada, Germany, and Italy. All country-pairs in which the industry from

home country was involved were coded as private sector.� In Australia, the entity engaging in AIJ projects had the nature of private-government

partnerships. We have coded them as government. Coding them as private sector projects

does not affect our results.

Table 1 describes projects in terms of the actor type from the home country (private sector and

government/NGO) and the project type in the host country (energy projects and sequestration

projects).

The incentives for developing countries to engage in AIJ projects are obvious. Energy

sector projects allow them to increase energy production and yet lower incremental pollution.

This is important because high levels of urban air pollution are one of the most pressing

environmental problems in many developing countries. Carbon sequestration projects could

lead to new revenues from their natural resources, such as forests, without destroying them.

The location drivers for developed countries are less obvious. While AIJs are likely to

be viewed as laboratories for future policy experiments, the location decisions are likely

to be driven by more concrete considerations. For example, home countries might want to

invest in countries without incurring huge costs to understand the local environment and

institutions. Prior knowledge of the host country would substantially lower the transaction

costs of conducting these policy experiments. For energy projects, home countries might

want to locate their projects in host countries that transfer airborne pollution to them. By

doing so, home countries might reduce their own pollution. And of course, location decisions

might be motivated by humanitarian as well as strategic considerations. This paper, therefore,

systematically examines the factors explaining home countries’ AIJ location decisions. The

next section identifies theoretical arguments that bear upon the location puzzle. The third

section describes the data and provides an empirical test of the analytical model along with

specification checks examining data disaggregated by actor type in the home country and

project type in the host country. The concluding section discusses implications of our results

for a broader analysis of international environmental cooperation.

Theoretical perspectives

The objective of the UNFCCC is to protect the atmosphere from being overused as a pollution

sink. It is extremely difficult (almost impossible, given that the UNFCCC is not enforceable)

for any country to prevent other countries from using the atmosphere as a pollution sink.

Therefore, if one country unilaterally reduces its use of the atmosphere as greenhouse gas

sink, there is no assurance that other countries will do the same. In addition to the “non-

excludability” issue, there is a “rivalry” problem as well: one country’s use of the atmosphere

as a sink limits the ability of others to do the same. These limits might not always be physical

Springer

Policy Sci (2006) 39:233–248 237

and occur in the short time, but political and evolve slowly. For example, the use of the

atmosphere as a sink for greenhouse gases predominantly by developed countries over the

last hundred years is now creating political pressures on China and India to restrain their

use of the atmosphere as a sink. These two characteristics – non excludability and rivalry,

pervasive in common-pool resources, create conditions for resource overuse and resource

degradation (Ostrom, 1990). In effect, the capacity of the atmosphere to absorb greenhouse

gases is a common-pool resource. Therefore, any attempt to protect this resource is beset

with collective action problems.

Beginning with Hardin’s “Tragedy of the Commons” (1968), there is an extensive literature

on how the ’commons dilemma’ may be mitigated. Virtually all contributors to the commons

debate agree that unilateral action is unlikely to work given that commons dilemma tends to

have the structure of an ‘n person prisoner’s dilemma’ game (Ostrom, 1990). In this situation

of interdependence (the final outcome depends not only on my action but actions of others as

well), a set of rules or institutions will need to be established that can curb resource overuse.

This is what the UNFCCC and eventually the Kyoto Protocol set to do. In the context of global

warming, unilateral actions may not suffice because no single country emits an overwhelming

proportion of the global emissions. Hence, the final outcome – the extent to which emissions

of greenhouse gases can be stabilized – depends on the actions of multiple actors, i.e., on

the effectiveness of the UNFCCC and the Kyoto Protocol. The collective action problems

are accentuated because of uncertainty about the extent and the geographic distribution of

impacts of global climate change as well as by the lack of sanctioning mechanisms for

countries that do not meet their emission reduction targets (Victor, 2001).

Given the common pool resource characteristics of the atmosphere, the obvious questions

are: why do we see variations in developed countries’ commitment to protect this global

common and how do they select their partners for bilateral AIJ projects? Regarding the

first question, the literatures on international cooperation (Victor, 2001) and common-pool-

resource governance (Ostrom et al., 2001) suggest that in the absence of an enforceable

regime, few countries will commit to curbing global climate change through any international

regime (Dolsak, 2001).2 Some sorts of selective incentives might create conditions for them

to follow through on their commitments.

AIJ projects (and eventually CDMS under Kyoto protocol) are a mechanism to create such

selective incentives. AIJs serve as policy laboratories to assess whether developed countries

can employ a low cost option of emission reductions by investing abroad in countries where

the marginal costs of emission reductions are lower in relation to home. While AIJ projects

did not offer emission credits, there was a possibility that in the future such credits might be

allowed. AIJ projects therefore serve as testing grounds for country-pairs to assess types of

projects around which cooperation might be feasible.

But where might such experiments get conducted? Or, where might the AIJ projects be

located? There seems no clear pattern explaining how home countries select host countries for

the AIJ projects. According to the data provided by the UN, in 1999, there was a concentration

of AIJ projects in Europe and in South America. Of 122 AIJ projects in 1999, 79 were located

in economies in transition, 29 in South America, 9 in Asia and 5 in Africa (UNFCCC, 1999).

Spatial concentration of these projects was a concern for the EU countries as well (European

Environment Agency, 2002). As Figure 1 indicates, some home countries tend to invest in

host countries that are in their immediate vicinity (solid arrows). These include Sweden with

2 They may, however, decide to re-label some of their current projects as “climate mitigation” projects. This isnot associated with any costs, but indicates the country’s or the industry’s commitment to the regime. Thereis a concern that some AIJ projects might be relabeled, business-as-usual projects (Michaelowa, 2002).

Springer

238 Policy Sci (2006) 39:233–248

Fig. 1 Illustration of location of investment for some AIJ projects

numerous AIJ projects in Estonia, Lithuania, and Latvia, and Japan with AIJ projects in

Asia. However, other developed countries such as France, the Netherlands and Norway have

undertaken AIJ projects that are not in their geographical proximity (dotted arrows).

Complex phenomena, AIJ location decision in our case, seldom have mono causal expla-

nations and a systematic study of several factors or drivers is required. We hypothesize that

two broad category of motivations, philanthropic and instrumental, are likely to influence the

location decisions.

If AIJ location decisions are influenced by the desire to (eventually) produce a global

public good (climate change mitigation), the AIJ location decisions will be influenced by

host country’s contribution to the global emissions. Developing countries that mine (and

burn) significant quantities of coal as well as countries with substantial forest cover will be

obvious choices in this regard. If a host country has a small impact on global climate change,

then AIJ investments in that country may have a marginal impact on overall reductions in

emissions or overall enhancements of carbon sinks. Thus, home countries may find these

countries less attractive for locating their AIJ projects.

Alternatively, AIJ projects might be viewed as foreign aid, not as policy experiments. While

philanthropy would still guide home countries’ location decisions, it would be unrelated to

their desire to mitigate global climate change. To test for the role of philanthropic but non-

climate change related motives, we test for two variables. The first one, GDP per capita,

captures the “need” for development aid in a developing country. If the home-country’s

AIJ location decisions are influenced by the desire to help the poor countries overseas –

irrespective of the recipient’s contribution to global climate change, we would expect a

negative regression coefficient for GDP per capita. Alternatively, home countries may want

to help developing countries but focus on the ones which serve their strategic interests. We

decided to examine the role of colonial ties because such ties capture the humanitarian as

Springer

Policy Sci (2006) 39:233–248 239

well as strategic aspects of “need.” If philanthropic cum strategic interests influence location

choices, we would expect home countries to invest in their former colonies. In some ways,

this is a recognition of and compensation for the colonial exploitation. Because many former

colonies maintain strong ties with their former colonizers, AIJ projects might further home

countries’ strategic interests as well. Our analysis, however, does not support this hypothesis.

Colonial relationships do not have statistically significant impact on the location of AIJ

projects. Because its exclusion did not affect our substantive results, the colonial ties variable

is not included in the final model.

The second set of factors focuses on purely instrumental reasons guiding location deci-

sions. If AIJ projects are policy laboratories, home countries would like to conduct these

experiments in countries where they have prior knowledge of local institutions and politics.

Otherwise, much effort would be required in acquiring local knowledge and the core issue,

whether AIJ projects are a cost efficient way of reducing emissions or enhancing sequestra-

tion, would get insufficient attention. How does one get to know the local context? In some

ways, prior exchanges would be helpful in the regard. These include bilateral trade as well

as aid given by the home country to the host country. With these exchanges, home countries

are expected to acquire knowledge about host country institutions and politics. Thus, the

transaction costs of establishing new projects, AIJ projects in our case, would be lower. This

familiarity is especially important because transaction costs for AIJ projects are substantial

(Powell, Lile, and Toman, 1997; Springer, 2003; Michealowa, 2002). To account for this

familiarity, our model includes trade flows between a host and a home country as well as aid

given by the home country to the host country.

Some other instrumental concerns might guide the location decisions. If home countries

receive pollution generated in host countries, the AIJs may serve as useful instruments (in the

short run or as platforms for projects in the long run) for home countries to reduce regional

air pollution. To illustrate, consider the situation in Asia or Europe. Observational data (Jaffe

et al., 1999) and models of global air pollution (Jaegle et al., 2003) indicate that air pollution

from China is getting transported to downwind countries such as Korea and Japan. Thus,

Japanese and Korean investors may have instrumental reasons to locate their AIJ projects

in China. Regional air pollution transfers modeled for Europe suggest that pollutants are

transferred by air currents from neighboring transitional economies to Scandinavian countries

(Barret et al., 1995). For example, source-receptor matrices calculated for European countries

suggest that Sweden receives 86% of its oxidized nitrogen from emission sources outside the

country (Berge et al., 1999). In such cases, if Scandinavian investors decide to locate their AIJ

projects in the Baltic or Central European countries, such AIJ location decisions are likely

to be motivated not only by global environmental concerns but by regional environmental

problems as well. If this logic holds, we would expect to see an important role of geography

in AIJ location decisions.3 This impact would be particularly relevant for energy projects as

they bear upon air pollution problems but less likely for the carbon sequestration projects

which do not have obvious down-wind or regional impacts.

3 The data available for 155 AIJ projects suggest that 63 of these 155 projects are also estimated to decreasenitrogen oxide emissions, hence local and regional air pollution. While AIJ emission reduction estimatesare uncertain and often poorly reported (Ott, 1998; Michaelowa, 1998, 2002), they are non-trivial. Moreimportantly, because AIJ projects might be viewed by the investors as “laboratories”(Michaelowa, 2002) forfuture projects that might lead to more significant reductions in carbon dioxide and nitrogen oxide emissions,the size of AIJ projects (and the potential reductions in emissions) is not likely to be important for theseinvestors.

Springer

240 Policy Sci (2006) 39:233–248

Table 2 Descriptive statisticsVariable Mean St. Dev. Min Max

Host GDPCap 4619.171 4109.903 497.6277 22231.93

Aid 16.93029 131.0914 0 5037.36

Trade 3.98e ± 08 2.52e ± 09 0 8.23e ± 10

Host coal 21.57553 113.8026 0 1190.38

Proximity 5824.323 2453.952 0 10812

Host forest 30.83319 23.56846 .003231 92.17577

Empirical model

This paper employs a logistic regression model to analyze factors impacting a home country’s

AIJ location decisions. The unit of analysis is a home-host country pair. The country-pairs

that had at least one AIJ project are coded as 1, those with no project as 0. We include all

AIJ projects from 1994 until 2002, the last year any AIJs were initiated. The source of the

data is the UNFCCC dataset on AIJ projects (UNFCCC, 2002d). As the data on project

characteristics are not of uniform quality, especially data on costs of the project and on the

level of implementation of the project,4 we focused only on presence of an AIJ project in a

country-pair. Because our dependent variable is dichotomous, we use a logistic regression

model (Long, 1997).

The data on independent variables are for 1990. Thus, our independent variables are prior

to the AIJ location decisions, most of which were undertaken in the first half of the 1990s.

For countries that did not exist in 1990 or were in political turmoil, we used data for the first

available year. This was 1993.

Our independent variables are of two types. For a detailed description of independent

variables, see Appendix A. First, we include the country-pair relationships that potentially

influence the transaction costs of doing business in host countries. As a proxy for transaction

costs, we measure aid and trade flows between a home and a host country in each country-pair

in years prior to the commencement of the AIJ projects. We use bilateral trade data published

by the OECD for years 1990 or 1993. Descriptive statistics of the independent variables are

presented in Table 2.

As a specification check, we also looked at sectoral trade data most relevant for specific

AIJ projects. For AIJ projects in the energy sector, we replaced total trade by home country’s

exports of electricity generation technologies and electric appliances to host countries. This

measure captures the home country’s presence in the host country’s market for technologies

that are most often targeted by the energy sector AIJ projects. For carbon sequestration

projects, we replaced total trade by exports of wood, paper, and cork by a host country

to a home country. This variable measures familiarity of the home country with the host

countries’ forest policies and suppliers, as well as the home countries’ potential domestic

consumer pressure for sustainable forestry practices. Relationships between trade and our

dependent variables were statistically significant for both total trade as well as sectoral trade.

The coefficients for other variables were also comparable in the two trade models. Because

4 Arguably, in a dynamic sense, poor implementation of AIJ projects might lead home countries to locatefuture AIJ projects elsewhere. Our sense is that this is a remote possibility. In addition, given that the unitof analysis is a country pair, one would have to develop a scheme of weighting projects, given that in acountry-pair year, there might be multiple projects. Most importantly, we are not aware of systematic data onproject implementation (across countries, over time) that might enable us to include project implementationas a covariate in our model.

Springer

Policy Sci (2006) 39:233–248 241

sectoral trade is highly correlated with aggregate trade (0.9), both types of trade measures

should not be included in the model. Hence, we have retained aggregate trade in the model.

In process of providing foreign aid, developed countries often seek to assess the ground re-

alities in developing countries. Often elaborate mechanisms are established to ensure that the

aid is used wisely. As a result, developed countries come to acquire considerable knowledge

about local institutions and politics. Thus, transaction costs involved in establishing and

managing AIJ projects are likely to be lower if home countries have provided aid to specific

host countries. To measure aid flows, we use the OECD dataset for year 1990. For the host

countries that did not exist in 1990 or were in political turmoil in 1990, we use the 1993 data.

Host country characteristics are also likely to influence AIJ location decisions. If AIJ

projects are viewed by a home country as instruments to create global public goods, then

home country investors may tend to locate AIJ projects in host countries that potentially

have high impact on global climate change. We measure this impact via host-country’s

annual coal production. By the same logic, home country investors interested in carbon

sequestration projects may tend to locate their AIJ projects in host countries with a substantial

forest cover (measured as percent of area covered by forest). This measure helps us to

investigate whether home countries’ location decisions for carbon sequestration projects are

influenced by opportunities offered in the host-country to avoid deforestation and to establish

agroforestry industry.

Home countries might be scrutinized not only for the potential for avoided deforestation,

but also for the potential for aforestation or forest growth. To put it simply, one could sequester

carbon either by ensuring that existing trees are not cut (avoid deforestation) or by growing

more trees (aforestation as well as forest growth). To control for host-country’s potentials for

aforestation, we examined a variable measuring the difference in the proportion of land under

forests between 1961 and 1993.5 The variable was not statistically significant in predicting

the location of AIJ carbon sequestration projects even in a univariate regression. Because its

exclusion does not affect our substantive findings, we have excluded it from the final model.

Finally, we investigate whether “coal production” effects are more salient for host countries

that are in home countries’ physical proximity. If home countries are motivated to reduce

regional pollution via the AIJ projects, then AIJ projects will tend to be located in host

countries that are up-wind from the home country. Unfortunately, modeling the exchange

of air masses between countries is complex and one cannot create unambiguous down-wind

measure. This is because in some months one country could be up-wind from another but in

other months, the situation could be reversed (Jaegle, 2005). In this paper, we therefore use

physical proximity to capture the effect of transport of pollution via air to other countries.

Clearly, proximity is likely to be more important when a nearby host country has a large

coal mining industry simply because this country is likely to burn coal, create pollution

which eventually will reach the home country. To capture this interactive effect, we included

an interaction term in our model. To maintain a meaningful interpretation of regression

coefficients of the original or lower order variables (proximity and coal production in a host

5 Clearly, this is a very crude measure of aforestation/forest growth potential in any country. While forestgrowth rates have been modeled for selected countries (LaMarche et al., 1984; Hasenauer et al., 1999; Maseraet al., 2003), we are only aware of a study by Niles et al. (2002) reporting comparable estimates of forestgrowth for 48 developing countries. While our dataset includes 129 countries, limiting the analysis to the48 developing countries for which the data are available would reduce the number of country-pairs from theinitial 2541 to only 936 and exclude about one third of sequestration AIJ projects from the analysis. When were-estimated the model under these restrictions, the model did not converge.

Springer

242 Policy Sci (2006) 39:233–248

Table 3 Unstandardizedcoefficients for logistic regressionof all AIJ projects

Variable Coefficient

Host GDPCap 0.0000381∗

(0.0000283)

Aid 0.00148∗∗

(0.000617)

Trade 5.61e–11∗∗

(2.46e–11)

Host Coal 0.0012484••

(0.0006986)

Proximity 0.0001865••

(0.000055)

Proximity1× Host Coal1 3.01e-07••

(0.000055)

Host Forest 0.0122467∗∗

(0.0054256)

Notes. Single variable test: ∗ p≤0.10, ∗∗∗p≤ 0.05, ∗∗∗p≤ 0.01(one-tail). Joint significance test:• p≤ 0.10, •• p≤ 0.05, •••p≤ 0.01(one-tail). N = 2541; Prob >

chi2 = 0.0000; Standard errorsare reported in parentheses

Table 4 Unstandardizedcoefficients for logistic regressionof AIJ projects, by type of project

Variable Energy AIJ projects Sequestration AIJ projects

Host GDPCap 0.0000275 0.0000781∗

(0.0000292) (0.000055)

Aid 0.0016159∗∗∗ 0.0005176•••

(.0006021) (0.0004374)

Trade 5.67e–11∗∗ 8.61e–11•••

(2.49e–11) (2.79e–11)

Host coal 0.0012603••• –

(0.0006938)

Proximity 0.000208••• –

(0.0000595)

Proximity1x host 3.59e–07••• –

coal1 (5.18e–07)

Host forest – 0.01982∗∗

(0.0105326)

Prob > chi2 0.0000 0.0007

Notes. Single variable test: ∗ p ≤0.10, ∗∗ p ≤ 0.05, ∗∗∗p ≤ 0.01(one-tail). Joint significance test:• p≤ 0.10, •• p≤ 0.05, ••• p≤ 0.01(one-tail). N = 2541. Standarderrors are reported in parentheses

country), we reparameterized the model by subtracting the means of the variables from each

of the two variables before creating the interaction term (Wooldridge, 2003:194).

Results and specification checks

Table 3 presents the results of the logistic regression analyses of AIJ projects between 1994

and 2002. In addition to analyzing AIJ location choices in the aggregate, we also examine

whether factors impacting location decisions vary across project types (sequestration and

energy, Table 4) and actor types (government and private sector, Table 5).

As hypothesized, trade and aid relationships are important drivers of AIJ location decision

across all types of projects (Table 4) and actors (Table 5). By reducing transaction costs,

previous exchanges enable home country investors to map out the terrain in the potential

host countries and provide confidence that their AIJ projects will not be hampered by local

idiosyncrasies. For one standard deviation increase in bilateral aid from a home country to a

Springer

Policy Sci (2006) 39:233–248 243

Table 5 Unstandardizedcoefficients for logistic regressionof AIJ projects, by type ofinvestor

Variable Government Projects Industry Projects

Host GDPCap 0.0000359 0.0000577∗

(0.0000353) (0.0000392)

Aid 0.0004047• 0.0016934∗∗

(0.0004079) (0.0006712)

Trade 2.16e–11• 6.78e–11∗∗

(2.82e–11) (2.68e–11)

Host Coal 0.0018326••• −0.0005746•

(0.0007099) (0.0014772)

Proximity 0.0002106••• 0.0001396•

(4.20e-07) (0.0000793)

Proximity1x Host Coal1 2.83e–07••• −4.83e-07•

(4.20e–07) (4.69e–07)

Host Forest 0.012892∗ 0.0128568∗∗

(0.006709) (0.0078728)

Prob > chi2 0.0003 0.0000

Notes. Single variable test: ∗ p≤0.10, ∗∗p ≤ 0.05, ∗∗∗ p ≤ 0.01(one-tail). Joint significance test:•p ≤ 0.10, •• p≤ 0.05, ••• p≤ 0.01(one-tail). N = 2541. Standarderrors are reported in parentheses

host country, the odds of this home country locating an AIJ project in this host country are

21% greater, holding all other variables constant. Similarly, for a standard deviation increase

in total trade in the country-pair, the odds of this country-pair engaging in an AIJ are 15%

greater, holding all other variables constant.

We also find that the interaction of physical proximity with coal has an important and

statistically significant effect on home countries’ location decisions (Table 3) across actors

(Table 5), specifically for AIJ energy sector projects (left panel of Table 4). For example, hold-

ing host countries’ coal production at its mean (about 20 million short tons per year, which is

about an equivalent of coal production of Colombia or Hungary), the odds of receiving an AIJ

project from a given home country by a host country A that is about 2400 miles closer to this

home country than another host country B, are about 60% greater. Thus, our results strongly

suggest that instrumental concerns regarding reductions in local/regional air pollution that

emanates abroad are driving AIJ location decisions rather than pure philanthropic reasons

pertaining to mitigation of global warming.

Our analysis suggests that home countries locate sequestration projects in host countries

depending on the extent of the area covered by forest. Holding all other variables constant,

the odds of a host country with about 10% larger area covered with forest (an equivalent

to Costa Rica with 40% of area covered with forest vs. Guatemala with about 30%) being

chosen for a sequestration AIJ project are about 20% greater.6

We also conducted a specification check by dropping an outlier that might be skewing

the results. This outlier is Sweden which has located AIJ projects in countries that are in its

physical proximity. Even when we exclude Sweden from the analysis, proximity continues

to have a statistically significant impact on location choice in the aggregate model (Table 6).

Holding coal production at its mean, the odds of a host country, which is about 2470 miles

closer to the home country than another host country, being engaged in an AIJ with this home

country are about 58% greater.

6 Geography is not a statistically significant driver (either alone or in interaction with forest cover) of thelocation of carbon sequestration projects. Given that proximity is not statistically significant even in theunivariate regression model, we did not include this variable in the final sequestration model reported in thepaper.

Springer

244 Policy Sci (2006) 39:233–248

Table 6 Unstandardizedcoefficients for logistic regressionof all AIJ projects, Swedenexcluded

Variable Coefficient

Host GDPCap 0.000036

(0.0000291)

Aid 0.0014673∗∗

(0.0006157)

Trade 5.76e–11∗∗

(2.50e–11)

Host coal 0.0013099•••

(0.000698)

Proximity 0.0001645•••

(0.0000553)

Proximity1x host coal1 4.26e–07•••

(5.04e–07)

Host Forest 0.0109973∗

(0.0055583)

Notes: Single variable test: ∗ p ≤0.10, ∗∗p ≤ 0.05, ∗∗∗p≤ 0.01(one-tail). Joint significance test:•p ≤ 0.10, ••p≤ 0.05, •••p ≤ 0.01(one-tail). N = 2420; Prob >

chi2 = 0.0000. Standard errorsare reported in parentheses

While our analysis suggests that there is a statistically significant relationship between a

host-country’s GDP per capita and the likelihood that a particular pair of home-host country

would have an AIJ project, the relationship is positive – in the direction opposite to the

hypothesized direction. This suggests that AIJ location decisions on the part of home country

are not influenced by the factors that typically drive their decisions regarding the developing

countries to which they will provide aid. Surprisingly, this relationship was significant for

private sector investors (right panel, Table 5), but not for government investors (left panel,

Table 5).

Conclusions

Regime literature has primarily examined conditions under which regimes get established.

Regime scholars have debated, given anarchy, what conditions facilitate collective action by

sovereign actors to achieve common goals that they unilaterally cannot achieve (Oye, 1986).

This literature has typically paid less attention to the mechanisms and policy instruments

by which regime goals are sought to be pursued and whether such instruments achieve their

stated goals.7 This paper contributes to this under-explored area by examining a concrete

policy measure that evolved from the UNFCCC regime. Unlike the much discussed Clean

Development Mechanism developed under the 1997 Kyoto Protocol, home countries invest-

ing in AIJ projects cannot earn emission reduction credits by virtue of their investment abroad.

Hence, it is puzzling as to why 12 developed countries have invested in 158 AIJ projects

in 42 developing countries and East European transition economies. The more interesting

puzzle is: how do the 12 developed countries decide where to locate their AIJ projects?

This paper examined both the instrumental and non-instrumental drivers of home coun-

tries’ AIJ location decisions. Because marginal costs – economic and political – of reducing

emissions may be lower in developing countries, the AIJ projects served as a policy labo-

ratory to assess whether such investments might be advantageous to both countries in the

event future regimes allowed emission credits from such bilateral projects. To conduct these

7 Exceptions include scholars that have begun to look at regime efficacy (Young, 1999; Miles et al., 2002).

Springer

Policy Sci (2006) 39:233–248 245

experiments with low transaction costs, our analysis suggests that home countries are likely

to locate AIJ projects in host countries with which they have had prior exchanges. Indeed

we find that prior trade and aid relationships with a home country improve the likelihood

of a host country getting AIJ projects from that country. Thus, instead of investing in host

countries where maximum pollution reductions (or carbon sequestration) might be possible,

home countries invest in locations where they can conduct their policy experiments at low

transaction costs.

Regarding energy projects, our key conclusion is that location decisions are driven by

home countries’ desire to reduce air pollution that they receive from abroad. Thus geography

– proximity of a host country to a home country – in interaction with host country’s coal

production, is a very important driver of location decision in AIJ energy sector projects.

Location of sequestration projects is impacted by the host country’s potential for avoiding

deforestation as well as by previous aid and trade patterns between a home and a host country.

Proximity is not important in this case.

The implications for regime design are obvious. Given that developed countries’ climate

change philanthropy is also influenced by instrumental reasons, when devising mechanisms

to implement any global environmental regime, policy makers should pay special attention

to regional pollution dynamics. Given the high costs of curbing the emission of greenhouse

gases, any global regime must offer tangible and excludable benefits to the signatories. While

in the case of AIJ, these benefits stem from reduction in regional and local pollution, other

types of benefits for the developed countries (that bear the largest proportion of the cost of

implementing global regimes) may be salient. The bottom line is that countries may not join

regime for philanthropic reasons alone; instrumental concerns may be very important. Those

seeking to enhance multilateral cooperation via international regimes should therefore pay

close heed to both the costs and benefits that regime creates for the key actors.

Appendix A: Description of independent variables

A.1. Aid

This variable measures total bilateral net aid (total official development aid) given by the

developed country to the developing country in the country pair in year 1990 (1993 for

countries that did not exist in 1990). Reported in millions US Dollars. Source: OECD, 2005.

A.2. Deforest6193

Reduction in the share of area covered with forests and woodlands between 1961 and 1993.

For countries that did not exist in 1961, such as the newly independent states that were

constituent parts of the USSR, Yugoslavia, and Czechoslovakia we estimated the data for

1961 assuming uniform forest change across the country. Measured in percent points. Source:

United Nations Environment Program, 2005.

A.3. ExpPoweGen

This variable measures exports of technology for electricity generation from the developed

country to the developing country in the dyad. For Belgium and the Netherlands, the Harmo-

nized System classification data for category 84 are used. For other countries, data for SITC

Springer

246 Policy Sci (2006) 39:233–248

rev. 3 category 71 are used. Source: OECD International Trade in Commodity Statistics,

2001.

A.4. ExpElectAppl

This variable measures exports of electric appliances from the developed country to the

developing country in the dyad. Source: OECD International Trade in Commodity Statistics,

2001.

A.5. Host coal

This variable measures domestic coal production in the host country. It includes “(t)he sum of

sales, mine consumption, issues to miners, and issues to coke, briquetting, and other ancillary

plants at mines. Production data include quantities extracted from surface and underground

mines, and normally exclude wastes removed at mines or associated preparation plants.”

Reported in million short tons of coal production in country in 1990 (1993 for countries that

did not exist in 1990). Source: US, Energy Information Administration, 2005.

A.6. Host coal1

Calculated by subtracting the mean of Host coal from Host coal.

A.7. Host forest (% area)

This variable measures the percentage of the area covered by forest in 1990 (1993 for countries

that did not exist in 1990). Source: World Bank, World Development Indicators, 2005.

A.8. Host GDPCap

Gross Domestic Product per capita in 1993, expressed in PPP, constant 2000 international

Dollars. Source: World Bank, World Development Indicators, 2005.

A.9. ImpWood

This variable measures imports of paper and paper products, cork and cork products, and

wood and wood products from a developing to the developed country in the dyad. For Belgium

and the Netherlands, the Harmonized System classification data for categories 44, 45, 47,

and 48 are aggregated. For other countries, data for SITC rev. 3 categories 24, 25, and 63 are

aggregated. Source: OECD International Trade in Commodity Statistics, 2001.

A.10. Proximity

A.10.1. Measured in miles

Proximity for a home country i and a host country j is calculated for each country-pairi j as

the difference between the maximum possible distance for the home country and the actual

distance between the capitals of the home and the host country in this pair. For example,

proximity for Australia and Philippines was calculated using the distance between Canberra,

Springer

Policy Sci (2006) 39:233–248 247

Australia and Rabat, Morocco (11,065 miles) as the maximum distance and the distance

between Canberra and Manila, Philippines (3905 miles).

Proximity country pairi j = Distance max for home countryi – Distance country-pairi j

Distance max for home countryi = max distance country-pairi j ; i is constant and j varies

from 1 to 129. i is home country ID. j is host country ID. Source: U.S. Department of

Agriculture.

A.11. Proximity1

Calculated by subtracting the mean of proximity from proximity.

A.12. Trade

This variable measures bilateral trade as reported by the developed country in the country pair

in year 1993. The data are reported in thousands of US Dollars. Source: OECD International

Trade in Commodity Statistics, 2001.

Acknowledgements Previous version of this paper was presented at the 2005 annual research conference ofthe Association for Public Policy Analysis and Management. We thank Matt Auer, Peter Hoff, Aseem Prakash,and two anonymous reviewers for their comments.

References

Barrett, K., Seland, Ø., Foss, A., Mylona, S., Sandnes, H., Styve, H., & Tarrason L. (1995). European trans-boundary acidifying air pollution. EMEP/MSC-W Report 1/95, July 1995.

Berge, E., Bartnicki, J., Olendrzynski, K., & Tsyro, S.G. (1999). Long-term trends in emissions and trans-boundary transport of acidifying air pollution in Europe. Journal of Environmental Management, 57,31–50.

Dolsak, N. (2001) Mitigating global climate change: why are some countries more committed than others?.Policy Studies Journal, 29, 414–436.

European Environment Agency. (2002). Exploring the ancillary benefits of the Kyoto Protocol for air pollutionin Europe. Copenhagen, Denmark: European Environment Agency.

Hardin, G. (1968). Tragedy of the commons. Science, 162, 1243–1248.Hasenauer, H., Nemani, R.R., Schadauer, K., & Running, S.W. (1999). Forest growth response to changing

climate between 1961 and 1990 in Austria. Forest Ecology and Management, 122, 209–219.Jaegle, L. (2005) University of Washington. Personal Communication.Jaegle L., Jaffe, D.A., Price, H.U., Weiss-Penzias, P., Palmer, P.I., Evans, M.J., Jacob, D.J., & Bey, I. (2003).

Sources and budgets for CO and O3 in the Northeast Pacific during the spring of 2001: results from thePHOBEA-II Experiment. Journal of Geophysical Research, 108, 8802.

Jaffe D.A., et al. (1999). Transport of Asian air pollution to North America. Geophysical Research Letters,26, 711–714.

LaMarche Jr. V.C., Graybill, D.A., Fritts, H.C., & Rose, M.R. (1984). Increasing atmospheric carbon dioxide:tree ring evidence for growth enhancement in natural vegetation. Science, 223, 1019–1021.

Long, S.J. (1997). Regression models for categorical and limited dependent variables. Thousand Oaks: SagePublications.

Masera, O.R., Garza-Caligaris, J.F., Kanninen, M., Karjalainen, T., Liski, J., & Nabuurs, G.J., Pussinen, A.,De Jong, B.H.J., & Mohren, G.M.J. (2003). Modeling carbon sequestration in afforestation, agroforestryand forest management projects: the CO2FIX V.2 aapproach. Ecological Modeling, 164, 177–199.

Michaelowa, A. (1998) Joint implementation-the baseline issue: economic and political aspects. Global En-vironmental Change, 8, 81–92.

Michaelowa, A. (2002). The AIJ pilot phase as laboratory for CDM and JI. International Journal GlobalEnvironmental Issues, 2, 260–287.

Miles, E.L., Underdal, A., Andresen, S., Wettestad, J., Skjaerseth, J.B., & Carlin, E. M. (2002). Environmentalregime effectiveness: confronting theory with evidence. Cambridge, MA: MIT Press.

Springer

248 Policy Sci (2006) 39:233–248

Niles, J.O. (2002). Potential carbon mitigation and income in developing countries from changes in useand management of agricultural forest lands. Philosophical Transactions: Mathematical, Physical andEngineering Sciences, 360, 1621–1639.

OECD. (2001). International trade by commodity statistics., Database Edition, http://new.sourceoecd.org/vl= 8591365/cl = 13/nw = 1/rpsv/statistic/s15 about.htm?jnlissn = 16081218. Accessed on July 16, 2005.

OECD. (2005). Destination of official development assistance and official aid – disbursements. in InternationalDevelopment Statistics (IDS). Table 2a, http://www.oecd.org/dataoecd/50/17/5037721.htm. Accessed onJuly 29, 2005

Ostrom, E. (1990). Governing the commons. Cambridge: Cambridge University Press.Ostrom, E., Dietz, T., Dolsak, N., Stern, P.C., Stonich, S., & Weber, E.U., eds. (2001). The Drama of the

Commons. Washington, D.C: National Academy Press.Ott, H. E. (1998). Operationalizing ’joint implementation’ organizational and institutional aspects of a new

instrument in international climate policy. Global Environmental Change, 8, 11–47.Oye, K., ed. (1986). Cooperation under anarchy. Princeton, NJ: Princeton University Press.Powell, M., Lile, R., & Toman, M. (1997). Assessing the constraints and opportunities for private-sector

participation in activities implemented jointly: two case studies from the US initiative for joint implemen-tation. Washington DC: Resources for the Future.

Reforest the Tropics. Inc. (2006). http://ji.unfccc.int/CallForInputs/BaselineSettingMonitoring/HBarres/ De-scription.doc. Accessed on March 20, 2006.

Springer, U. (2003). Can the risks of the kyoto mechanisms be reduced through portfolio diversification?evidence from the Swedish AIJ Project. Environmental and Resource Economics, 25, 501–513.

United Nations Environment Program. (2005). Global environment outlook, Forest and Woodland Percent ofLand Area Table, http://geodata.grid.unep.ch/results.php. Accessed on August, 6, 2005.

UNFCCC. (1999). Activities implemented jointly under the pilot phase’, FCCC/SB/ 1995/5/Add.1 14. GE.99. http:/www.ccsr.u-tokyo.ac.jp/unfccc2/pdfs/unfccc.int/resource/docs/1999/sb/05a01.pdf. Accessed onFebruary 15, 2006.

UNFCCC. (2002a). Uniform reporting format: activities implemented jointly under the pilot phase: jelgavaenergy efficiency. http://unfccc.int/kyoto mechanisms/aij/activities implemented jointly/items/1779.phpAccessed on November 10, 2005.

UNFCCC. (2002b). Uniform reporting format: activities implemented jointly under the pilot phase:The model project for energy conservation in electric furnace used for ferro-alloy refin-ing. http://unfccc.int/kyoto mechanisms/aij/activities implemented jointly/items/1846.php Accessed onNovember 10, 2005.

UNFCCC. (2002c). Uniform reporting format: Activities implemented jointly under the pilot phase: city ofdecin: fuel switching for district heating. http://unfccc.int/kyoto mechanisms/aij/activities implementedjointly/items/1730.php. Accessed on November 10, 2005.

UNFCCC. (2002d). Activities implemented jointly. http://unfccc.int/kyoto mechanisms/aij/activities implem-ented jointly/items/2094.php. Accessed on September 6, 2005.

UNFCCC. (2005). Activities implemented jointly under the pilot phase: United Nations convention on cli-mate change. http://unfccc.int/kyoto mechanisms/aij/activities implemented jointly/items/2307.php. Ac-cessed on November 10, 2005.

United states department of agriculture. Great circle distances between capital cities. http://www.wcrl.ars.usda.gov/cec/java/capitals.htm Accessed on March 7, 2006.

Victor, D. (2001). The Collapse of the kyoto protocol and the struggle to slow global warming. Princeton:University Press

Wooldridge, J.M. (2003). Introductory econometrics: a modern approach. Second edition. Stamford, CT:Thomson.

World Bank Group (2005). World development indicators, http://devdata.worldbank.org/dataonline/Accessedon August 20, 2005.

Young, O., ed. (1999). The Effectiveness of international environmental regimes: causal connections andbehavioral mechanisms. Cambridge, MA: MIT Press.

Springer