Subnational Climate Justice? Adaptation Finance Distribution and Climate Vulnerability

World Development Vol. 58, pp. 130–142, 2014 Ó 2014 Elsevier Ltd. All rights reserved. 0305-750X/$ - see front matter www.elsevier.com/locate/worlddev

http://dx.doi.org/10.1016/j.worlddev.2014.01.014

Subnational Climate Justice? Adaptation Finance Distribution and Climate Vulnerability SAM BARRETT * Trinity College, Dublin, Ireland Summary. — This article investigates subnational adaptation finance distribution within Malawi. Malawi is highly climate vulnerable and a significant per-capita recipient of adaptation finance. This empirical study models distribution dynamics through “need” (climate vulnerability) and “government interest” (patronage). Results indicate those areas most in need receive relatively little finance. Rather, donor utility and the ability to absorb capital offer the most persuasive explanations for distribution across the state. These findings suggest that the distribution of adaptation funds do not support the larger goal of climate justice. Ó 2014 Elsevier Ltd. All rights reserved. Key words — climate justice, adaptation finance, climate vulnerability, Malawi

1. INTRODUCTION

Change Convention (UNFCCC) states “developed country parties. . .shall. . .assist the developing country parties that are particularly vulnerable to the adverse effects of climate change in meeting costs of adaptation to those adverse effects”. These perspectives are applications of Rawls (1972) “difference principle”: if a decision-maker has no information regarding their position within a society, they will distribute resources to the most disadvantaged groups. This article therefore contributes to the empirical literature that operationalizes justice or inequity through adaptation finance (Barrett, 2013b; Ciplet et al., 2013; Stadelmann, Persson, Ratajczak-Juszko, & Michaelowa, 2013). Adaptation finance research typically focuses on international distribution of climate funds (Ackerson et al., 2011; Stadelmann et al., 2013). Yet, inter-state distribution is but the first stage in a multi-scalar dissemination process to vulnerable communities. No research to date explores the subnational allocation of adaptation finance, or how governments within developing states are interested agents in the process. Rather, subnational distribution research investigates issues of food aid (Clay, Molla, & Habtewold, 1999; Jayne, Strauss, Yamano, & Molla, 2001; Jayne et al., 2003); natural disaster response (Besley & Burgess, 2002; Morris & Wodon, 2003; Francken, Minten & Swinnen, 2012; Takasaki, 2011); and education grant (Reinikka & Svensson, 2004) and total aid allocation (Zhang, 2004). Each study determines the role of equity and government interest in resource allocations to administrative units (districts, provinces, towns, and villages). This article focuses on adaptation finance distribution within Malawi, offering the first such subnational analysis and bridging the scales between central and district level government (Barrett, 2013a). The article provides explanations in accordance with the following questions:

The inverse relationship between climate risk and responsibility is the basis of climate justice debates (Adger, Paavola, Huq, & Mace, 2006; Ciplet, Roberts & Khan, 2013; Roberts & Parks, 2007). Since 1960, most least-developed states emitted less than 115 per capita tons of carbon, relative to 1.6–2.7k tons in many developed states (World Bank, 2013). But least-developed states experience far greater consequences from acute exposure to physical hazards (Busby, Smith, & White, 2011; Busby, Smith, White, & Strange, 2013; Maplecroft Index), including flooding (Lopez-Marrero, 2010; Mustafa, 1998), droughts (Eriksen & Lind, 2009; Stringer et al., 2009) and storms (Fazlul & Nobuo, 2008); all of which result in disproportionate losses. Yet, climate risk is determined by factors other than physical impact, including poverty rates, demographic characteristics, and governance (Blaikie, Cannon, Davies, & Wisner, 1994; Wisner, Cannon, Davies & Blaikie, 2004). This is most apparent in Africa, where underdevelopment results in disproportionate adverse social, economic, and physical security consequences, raising inequity within and across countries (Markandya, 2011; Mendelsohn, Morrison, Schlesinger, & Andronova, 2000). The climate justice debate investigates compensatory actions 1 given by developed states to address such inequalities (Ciplet et al., 2013; Clark, 2012; Hulme, O’Neill, & Dessai, 2011). Adaptation finance activities 2 are quantifiable compensations for the inequalities of climate change (Barrett, submitted for publication), including funds for livelihood diversification in drought prone areas, water conservation, flood mitigation, and climate-related disaster management. By following funds down to local-level implementation, the “climate justice” framework (Barrett, 2013a) identifies distribution paths to the most vulnerable states, districts, and communities: international adaptation finance transfers reveal whether the most climate vulnerable states are receiving assistance; subnational adaptation finance activities determine whether the most climate vulnerable districts are recipients. The climate justice imperative is whether vulnerability is the main determinant (Paavola & Adger, 2006, p. 604); vulnerable urban stakeholders are recipients (Ayers, 2009); and with equitable allocation to the most vulnerable countries (Bird, Brown, & Schalatek, 2011, p. 6). Indeed, the United Nations Climate

* This research is funded by the Irish Research Council for Humanities and Social Sciences, the Department of Geography in Trinity College and the Climate Change and African Political Stability Project at University of Texas - Austin. I would like to thank Ashley Moran, Mesfin Bekalo, Anna Davies, Justin Baker, Josh Busby and Catherine Boone who have all helped this project along at various stages. Final revision accepted: January 18, 2014. 130

SUBNATIONAL CLIMATE JUSTICE? ADAPTATION FINANCE DISTRIBUTION AND CLIMATE VULNERABILITY

Is climate vulnerability the primary determinant in adaptation finance distribution? Is government interest the dominant factor in adaptation finance distribution? Distribution to the most climate vulnerable districts will indicate the continuation of climate justice as a multi-scalar process (Barrett, 2013a). Yet, the broader subnational literature suggests other factors determine fund allocation. In particular, government interest and ethno-regional patronage drive resource allocations (Clay et al., 1999; Jayne et al., 2001; Reinikka & Svensson, 2004) and macro-theories of African politics emphasize the structural dynamics of ethnicity and patronage in goods provision (Azam, 2001; Posner, 2005). Therefore, the analysis applies a derivative of the “recipient need” and “government interest” model to adaptation finance distribution within Malawi. The results indicate high physical vulnerability is a driver of distribution, but high socioeconomic vulnerability displays a negative relationship; government interest also shows a negative relationship, which may be explained by programmatic policymaking in the latter years of the sample; instead, donor utility and district absorption offer the most persuasive explanations. Adaptation finance distribution arrives in districts with sufficient capacity to use assistance productively and where aid networks are established. The poorest, most marginalized, and climate vulnerable districts receive the least adaptation finance within Malawi. The article is divided into six sections: Section 2 provides an overview of climate justice and general subnational distribution literatures, introduces adaptation finance and its assimilation into existing distribution models; Section 3 outlines the methods of data isolation for broad and narrow samples; Section 4 outlines the model strategy; Section 5 presents the findings of the analysis; and Section 6 concludes with the climate justice implications of the research. 2. CURRENT LITERATURE This section outlines the climate justice literature, and situates adaptation finance data within existing studies of subnational resource distribution. New adaptation finance data assists in the investigation of questions previously unanswered. (a) Climate justice This article investigates the subnational distribution of adaptation finance, and moreover, contributes to an inter-disciplinary debate on climate justice and ethics. The literature to date is primarily philosophical. Several clusters of philosophical literature address cosmopolitan justice (Caney, 2006; Maltais, 2008) and intra- and intergenerational justice (Arler, 2001; Green, 1977), and applied research evaluates market mechanisms (Caney, 2010; Page, 2011). Other conceptually-based studies investigate justice in the values, knowledge, access, and property rights of the global forest regime (Forsyth & Sikor, 2013); and policy debates are beginning to establish empirical as well as conceptual bases for including justice in climate finance negotiations (Anderson, 2013) An empirical literature now addresses justice or equity issues, including means assessments of climate vulnerable localities (Dulal, Brodnig, Thakur, & Green-Onoriose, 2010); comparative studies of inequitable climate impacts (Griniski et al., 2012); justice implications of climate policies (BeymerFarris & Bassett, 2012; Mercer, Perales, & Wainwright, 2012; Suiseeya & Caplow, 2013); and equity-related benefits

131

of mitigation actions (Garg, 2011). Some studies quantify climate justice compensation: Roberts and Parks (2007) use environmental treaty ratification as a proxy measure; Stadelmann et al. (2013) focus on international distribution of the Adaptation Fund to find higher climate vulnerability associated with lower rates of adaptation finance distribution. To date, no research examines subnational adaptation finance distribution as a justice issue. (b) Subnational distribution A diffuse literature quantifies subnational resource distributions relating to foreign aid or government programs. These address topics such as food aid (Clay et al., 1999; Jayne et al., 2001, 2003); government responsiveness to disasters (Besley & Burgess, 2002); local capture of education grants, and related resources (Francken, Minten, & Swinnen, 2009a, 2009b; Reinikka & Svensson, 2004); infrastructure and public investments (Moser, 2008; Paxson & Schady, 2002); total aid (Zhang, 2004); and natural disaster assistance (Morris & Wodon, 2003; Francken et al., 2012; Takasaki, 2011). The aggregated findings indicate resources are allocated according to: (a) higher need; (b) government-location political linkages; (c) existing aid networks; and (d) high absorptive capacity (Jayne et al., 2001; Reinikka & Svensson, 2004). As yet, no large-n research models the determinants of subnational adaptation finance distribution. (c) Introduction of adaptation finance Adaptation finance navigates similar bureaucratic channels to other funds, but the final recipients and project objectives differ. Adaptation finance originates as bilateral or multilateral international financial transfers to developing states. There are multiple sources, but transfers are often made through dedicated climate change funds such as the Adaptation Fund, or as is most common, arrive as standard ODA (Official Development Assistance). 3 Analogous to other funds, adaptation finance arrives in either the Ministry of Finance of recipient governments or government-appointed implementing agencies. The next stage involves distribution to district administrations, 4 which manage intra-district implementation to village authorities via committee systems and multiple levels of traditional leaders. Once adaptation finance is implemented as a village level intervention, the differing objectives from ODA become apparent. Adaptation finance is designed to formally institutionalize the process of climate adaptation and resilience. Ayers and Huq (2009) identify this behavioral process as “the adjustment in natural or human systems in response to actual or expected climatic stimuli or their effects, which moderate harm or exploit beneficial opportunities”. Projects include livelihood diversification strategies, water management in surplus/deficit areas, drought resistant crop cultivation, and preand post-disaster management (i.e., disasters with an association to the direct and indirect effects of climate change) to name but a few. In Malawi, adaptation finance distribution navigates the same bureaucratic structures of central and local government explained previously. Again in accordance with the generic explanation above, the main difference is the specific objectives and climate vulnerable actors adaptation finance is designed to address/assist. As a consequence, adaptation finance requires the construction of a new model to capture explanatory dynamics, and in particular, a different conception of need.

132

WORLD DEVELOPMENT

(d) Climate vulnerability and government interest model

3. DATA

Need perspectives recall moral and humanitarian arguments about disproportionate resource distribution to disadvantaged groups (Chandrasekhar, 1965; Rawls, 1972). Some subnational distribution literature finds greater recipient need a driver of fund allocation (Morris & Wodon, 2003). If need determines distribution, climate vulnerable districts will consistently receive adaptation finance. Therefore:

This section first explains the case selection methodology, and contextualizes Malawi’s historical vulnerability to climate change; second, it outlines the narrow (climate-oriented) and broad (capacity development) adaptation finance-dependent variables; and thirdly it describes the independent variables.

H1. The higher the level of climate vulnerability within recipient districts, the greater the distribution of adaptation finance.

Malawi demonstrates a relationship between adaptation finance distribution and climate vulnerability (Barrett, submitted for publication). The multi-scalar climate justice framework follows money to highly climate vulnerable places with high per capita adaptation finance (Barrett, 2013a). Malawi is third ($5.60) and first ($4.95) highest per-capita African recipient of adaptation finance within broad and narrow categorizations respectively 5 (Barrett, submitted for publication); further, mean calculations of the Maplecroft index indicate Malawi as the most climate vulnerable mainland state in Africa (6.89). Malawi has a history of droughts, dry spells, flooding, storms, forest fires, and other climate-related hazards (Department of Environmental Affairs Malawi, 2006; Davies, Pollard, & Mwenda, 2010; Mijoni & Izadkhah, 2009). In particular, erratic and shortened rainy seasons are common (Phiri & Saka, 2008). Physical events interact with social and economic issues, such as reliance on rain-fed agriculture, limited livelihood options, disease, weak infrastructure, and general poverty (Benson, 2002; Chidanti-Malunga, 2011). Malawi also provides access to government records and which allows the geo-coding of recent aid activity (Weaver & Peratsakis, 2011). The Climate Change and African Political Stability Program (CCAPS) partnered with AidData and the Ministry of Finance in Malawi to locate the destinations of all aid activities from 2000 to 2010 (Peratsakis, Powell, Findley, Baker, & Weaver, 2012). This includes 27 bilateral and multilateral donors to Malawi, such as the World Bank, African Development Bank, the Norwegian government, the Japanese government, and the Department for International Development (DIFD) in the United Kingdom. 6 In addition, CCAPS developed an adaptation finance isolation methodology assisting the extraction of adaptation finance activities from the larger body of aid data (Weaver, Baker, & Peratsakis, 2012). Unprecedented access to government project documents has enabled tracking of aid activity classified as adaptation finance. 7 What follows describes adaptation finance activities developing capacity for resilience against climate risk (broad), and those climate-oriented in terms of enhancing adaptive capacity (narrow) (Figure 1).

Government interest offers the main alternative explanation, focusing on the Government of Malawi as director and supervisor of funds. In particular, African patronage theories suggest districts are rewarded for political support through resource distribution (Azam, 2001; Posner, 2005). The distribution literature finds districts rewarded for political loyalty in the context of Ethiopian food aid (Jayne et al., 2001) and World Bank loans to China (Zhang, 2004). Therefore: H2. As indicators of government interest increase within recipient districts, adaptation finance increases. There is a range of additional explanations. The most common is donor utility, which accounts for disproportionate aid distribution to districts with an established aid network, or as is sometimes termed, path dependency (Francken et al., 2009a, 2009b). This facilitates cost minimization and accessibility for donors (Francken et al., 2009a, 2009b; Francken et al., 2012) (Table 1). Therefore: H3. The higher the level of donor utility, the higher the level of adaptation finance allocation. Finally, absorptive capacity, media access, and agriculture may influence allocation. First, research finds income positively related to resource distribution (Reinikka & Svensson, 2004). Funds may be dispersed to those with adequate means of operationalizing funds into productive use. This is positively related to socio-economic indicators, and as such, offers a counter explanation if indicators show a negative relation to climate vulnerability. Second, governments tend to favor politically engaged citizens, who are likely to vote, and what’s more, vote for candidates advancing their interests (Francken et al., 2012). Finally, adaptation finance may simply be allocated to districts heavily engaged in agriculture, as opposed to the climate vulnerable. This controls for agricultural areas with low climate vulnerability attracting finance due to their productive characteristics.

(a) Case selection

(b) Dependent variable I: climate-oriented “Climate-oriented” is adaptation finance narrowly defined. This sample indicates the “motive or intent of the activity is

Table 1. Hypothesis, argument, and expectations Hypothesis

Argument

Expectation

H1: Climate Vulnerability

Normative

H2: Government Interest

Patronage

H3: Donor Utility

Efficiency

Positive and significant relationship between adaptation finance distribution and indicators of climate vulnerability Positive and significant relationship if H1 is rejected and government allocates to political supporters Positive and significant relationship if H1 is rejected (marginalized and exposed locations more costly to access), but may correspond with H2

SUBNATIONAL CLIMATE JUSTICE? ADAPTATION FINANCE DISTRIBUTION AND CLIMATE VULNERABILITY

133

crop production, seed supplies, afforestation, disaster reduction, and community participation (see Appendix A). This is a generalizable exercise sampling for the purposes of distributional analysis with relative counts of activities within districts and their determinants; this represents the known universe of adaptation finance within Malawi during 2000–10. The framework is applicable to other countries with geo-referenced adaptation finance data. 10

Figure 1. Adaptation finance distribution within Malawi 2000–2010.

framed by a changing climate, either past, present, or future” (Weaver et al., 2012, p. 5). The objective is vulnerability and risk reduction within the context of climate change, by enhancing adaptive capacity of human or natural systems, 8 including climate-proofing agriculture, disaster prevention, and irrigation (see Appendix A). (c) Dependent variable II: capacity development Projects may address the direct and indirect effects of climate change 9 and yet retain no explicit statement that the funds are climate related. Data collection requires maximum clarity on demarcations between samples, due to the endogeneity of socio-economic drivers to climate vulnerability and adaptation (Wisner et al., 2004). The exercise builds on the following: Roberts and Peratsakis (2011) sample DIFD’s aid transfers by human coding each project to determine degrees of “fit” to 19 categorizations of adaptation and resilience; Ackerson et al. (2011) conduct a similar operation using five definitions in operation within funding or research bodies (OECD, World Resources Institute and the Least Developed Countries project). Adaptation finance in this broader sense is termed “capacity development”. Capacity development indicates the project “does not have a climate-oriented motive, yet provides climate resilience. . .this separates capacity development from general development” (Weaver et al., 2012, p. 5). Capacity development adaptation finance reduces vulnerabilities of human or natural systems to effects of climate change risks through greater resilience to climate change. This includes activities for water supply/sanitation, agricultural water resources, food

(i) Explanation: climate vulnerability Climate vulnerability indicates need within the context of climate change. Research differentiates geographic locations by vulnerability through socio-economic and physical factors (Busby et al., 2011; Maplecroft Index Wheeler, 2011). The main shortcomings are a lack of time-varying measures and interpreting aggregated indices. This investigation uses three interpretable time-series vulnerability indicators that (a) maximize explanation of the various dimensions of vulnerability (exposure, sensitivity, and adaptive capacity); and (b) are suitable for the Malawi context: Physical exposure is operationalized through the Environmental Indications and Warnings Project (EIW) GIS dataset from the Global Climate Change Research Program (Global Climate Change Research Program, 2011a, 2011b, 2011c). 11 Information relates to freshwater surplus and deficit throughout the duration of the sample at 3, 5, and 10 years, capturing drought, dry spells, and flood as dominant physical hazards in Malawi (Department of Environmental Affairs Malawi, 2006; Davies et al., 2010; Mijoni & Izadkhah, 2009). 12 EIW raster data allows for the calculation of district scores. In the interests of interpretability, the model uses the variable “EIW”, which is a dummy variable indicating “1” if the district value is above the mean, and “0” if otherwise. Socio-economic sensitivity to climate impact is measured through infant mortality. This research uses the EIW and Maplecroft methodology to include data on infant mortality from the National Statistics Office of Malawi. “Quart_Inf_Mort” is a categorical variable (1–4) dividing infant mortality rates into quartiles in the interests of interpretability. Life expectancy is an indicator of adaptive capacity within Maplecroft and EIW methodologies. This research applies district level life expectancy data from the National Statistics Office of Malawi. The variable “Life_expect” standardizes above mean values with z-scores, while leaving those below with zero, to account for uneven distributions and to assist interpretation. Table 2 shows how adaptation finance activities correspond with levels of climate vulnerability. Columns indicate adaptation finance activities are consistently given to districts with high physical, but low socio-economic climate vulnerability. (ii) Explanation: government interest For government interest, the study includes the proportion of each district that voted for presidents Baliki Muluzu and Bingu Mutharika during 2000–10 in the prior election. 13 Muluzu was president from 2000 to 2004, Mutharika from 2004 to 2010, with years 2009 and 2010 corresponding to changes in support from the re-election of Mutharika in 2009. Data are collected from the African Elections Database, the Malawi Electoral Commission and Malawi SDNP information services. The variable “Vote_President” is a standardized z-score of values above the mean, while leaving those below with zero. Manipulations are made due to uneven distributions and ease of interpretation.

134

WORLD DEVELOPMENT Table 2. Adaptation Finance Activities and Climate Vulnerability Indicators

Climate vulnerability Climate oriented EIW Life expectancy Infant mortality Capacity development EIW Life expectancy Infant mortality

Low

Moderate

High

Highest

MNA CC MNA CC MNA CC

0.152 (0.06) 0.366 (0.16) 0.316 (0.17)

0.069 (0.00) 0.500 (0.14) 0.522 (0.30)

0.342 (0.09) 0.169 (0.32) 0.014 (0.13)

0.463 (0.17) 0.011 (0.06) 0.169 (0.10)

MNA CC MNA CC MNA CC

2.680 (0.08) 2.492 (0.13) 2.075 (0.10)

1.263 (0.03) 1.958 (0.16) 6.164 (0.38)

3.410 (0.00) 6.779 (0.35) 1.188 (0.43)

3.623 (0.20) 0.773 (0.07) 1.174 (0.01)

MNA = mean number of activities; CC = correlation coefficient.

Table 3. Adaptation finance activities and government support Government support Climate oriented Capacity development

MNA CC MNA CC

Low

Moderate

High

Highest

0.253 (0.10) 3.946 (0.14)

0.366 (0.24) 3.014 (0.06)

0.171 (0.06) 1.985 (0.22)

0.228 (0.10) 1.914 (0.03)

MNA = mean number of activities; CC = correlation coefficient.

Table 3 shows how adaptation finance activities correspond to levels of government support. Contrary to other studies, columns indicate that allocations are consistently made to districts with lower government support. (iii) Explanation: donor utility The relative cost to donors providing adaptation finance matters when deciding allocations. This study avails of the larger CCAPS dataset on geocoded ODA activities (Weaver & Peratsakis, 2011) and includes activities for general development aid within each district during 2000–10. The variable “ODA” counts activities by district-year. (iv) Other explanations Adaptation finance may be dispersed to districts with the means to productively use resources. Absorptive capacity is determined as follows: (a) if socio-economic vulnerability indicators of life expectancy and infant mortality are negatively related to distribution; and/or (b) if the proportion of the district that has a television displays a positive relationship with finance distribution. These indicators are related to poverty (poverty and owning a television: 0.41; poverty and life expectancy: 0.10; poverty and infant mortality: 0.42). Data on television ownership are gathered by clustering Demographic and Health Surveys (2000, 2004, 2010). The variable “Tele_owner” is a standardized z-score of above mean values,

while leaving those below with zero. Manipulations are made due to uneven distributions and ease of interpretation. Table 4 shows how adaptation finance activities correspond to poverty levels. Columns indicate the poorest receive the least activities and is most pronounced with climate-oriented activities. To account for media access driving allocation, the study interprets the proportion of television ownership (Tele_owner) as an indicator. Second, adaptation finance activities may be allocated to districts heavily engaged in agriculture. This study uses the proportion of agricultural engagement from clustering DHS surveys (2000, 2004, 2010) at the district level. The variable “Agri_engage” is a standardized z-score of values above the mean, while leaving those below with zero. Manipulations are made due to uneven distributions and for the purposes of interpretation. Finally, population is an inconsistent predictor that is highly correlated (0.45) with ODA and thus its influence on the model is covered by inclusion of the latter (Table 5). 4. MODEL STRATEGY In a panel setting, the relationship between adaptation finance [expressed in this section as Climate Finance for Adaptation (CFfA)] distribution and the main determinants take on the following econometric expression:

Table 4. Adaptation finance activities relative to poverty within Malawi districts Poverty Climate oriented Capacity development

MNA CC MNA CC

Poorest

Poor

Moderate

Rich

0.181 (0.24) 2.712 (0.07)

0.212 (0.32) 3.015 (0.05)

0.257 (0.13) 2.803 (0.10)

0.483 (0.04) 2.919 (0.00)

MNA = mean number of activities; CC = correlation coefficient.

SUBNATIONAL CLIMATE JUSTICE? ADAPTATION FINANCE DISTRIBUTION AND CLIMATE VULNERABILITY

135

Table 5. Variables and descriptive statistics Variable EIW Life_expect Quart_Inf_Mort Vote_President Agri_engage ODA Tele_owner

Observations

Mean

Standard dev.

Min

Max

286 286 286 286 286 286 286

0.587 0.372 2.461 0.435 0.424 2.276 0.348

0.493 0.756 1.141 0.597 0.180 2.325 0.793

0 0 1 0 0 0 0

1 3.542 4 1.732 0.869 15 5.295

CFfAit ¼ b0 þ b1 Cvit þ b2 Giit þ b3 Duit þ b4 Abit þ b5 Agit þ it ð1Þ where CFfA is the amount of activities to district i during period t, Cvit is the average climate vulnerability within the district; Giit, is government support in district i during period t; Duit, is ODA activity within district i during period t; Abit, is absorptive capacity within district i during period t; and Agit, is the proportion of district i engaged in agriculture during period t; and eit is the constant term capturing time invariant factors (see Appendix B for expectations of variable movements). (a) Regression specification The empirical strategy is dependent on inter-district variation to identify relative impacts of measures for recipient need, government interest, and other explanatory factors. Variation is expressed as two dependent variables: (1) a dichotomous variable indicates whether a district receives a CFfA activity within a given year; (2) a count variable displays the number of CFfA activities a district receives within a year. (i) Dichotomous variable The study uses a logit regression model to analyze the probability of receiving a CFfA activity (0, 1) as a function of the explanatory factors and vector of unknown parameters 14:   Pt Lt ¼ ln 1  Pt ¼ b0 þ b1 Cvt þ b2 Git þ b3 Dut þ b4 Abt þ b5 Agt

ð2Þ

whereby Lt refers to the time-series logit function, Pt is the probability of a district receiving an CFfA activity within a given year, b0 is the intercept from the linear regression equation, and b1, . . ., b5 are the regression coefficients multiplied by the values of the predictors. (ii) Count variable The count variable requires application of the negative binomial model, due to violation of the equi-dispersion assumption within Poisson models. Therefore: P fy t ¼ yjxt g ¼ where

Cðh þ yÞ h ry ð1  rt Þ ; Cðy þ 1ÞCðhÞ t

y ¼ 0; 1; 2; . . . ;

ð3Þ

log kt ¼ b0 þ b1 Cut þ b2 Git þ b3 Dut þ b4 Abt þ b5 Agt ð4Þ

and

rt ¼

kt h þ kt

ð5Þ

whereby yt is the count of CFfA activities within a given year, and y is the realized value of the variable. This is a multiparameter model with rt as the time-series dispersion parameter that applies log kt as the mathematical equivalent of kt being the base of the natural logarithm to the power of the intercept, plus the slopes and multiplied by regressors.

5. RESULTS (a) Need model Across the models physical and socio-economic factors of climate vulnerability diverge. Districts experiencing physical vulnerability from flood and drought receive more adaptation finance activities; the mean to maximum value yields between 7.30% (model 1) and 11.07% (model 4) greater chance of receiving funds within a given year. Likewise, the number of adaptation finance activities rise as physical vulnerability increases from mean to maximum values, corresponding to an increase in activities from 0.139 to 0.397 (model 7) and from 2.10 to 3.28 (model 10). Yet socio-economic drivers are negatively related to adaptation finance distribution. As infant mortality moves from the 2nd to 3rd quartile (moderate to high vulnerability), the probability of districts receiving climate-oriented activities falls 16.55% (model 2) and 48.96% for capacity development funds (model 5); likewise, the same shift in infant mortality decreases the number of climate-oriented and capacity development adaptation finance activities from 0.519 to 0.021 (model 8) and 5.784 to 1.295 (model 11), respectively. Further, as life expectancy increases from mean to maximum values, the probability of a district receiving a climate-oriented activity rises from 11.37% to 27.46% (model 3). Adaptation finance activities go to districts with low socioeconomic climate vulnerability but high physical vulnerability. At issue here is the primacy given to socio-economic factors within climate vulnerability frameworks, and their centrality to the aggregated concept of climate risk. Physical hazards represent only a small part of overall climate risk and socioeconomic factors are primary drivers of sensitivity, exposure, and low adaptive capacity (see weighting within Maplecroft Index and Busby et al. (2011)). From the founding vulnerability texts of Blaikie et al. (1994) and Wisner et al. (2004), the belief is that experiencing hazards does not necessarily increase climate risk. The relatively wealthy have sufficient resources to protect themselves from flood, drought, storms, or other direct and indirect effects of climate variability and change. Rather, to suffer adverse consequences, communities must be in a disadvantageous economic, social, and political position. Hence, districts at high climate risk are disproportionately low recipients of adaptation finance activities within Malawi.

136

WORLD DEVELOPMENT

The findings substantiate Stadelmann et al. (2013) now on the subnational scale and disappoint aspirations of activists, academics, and policymakers. Despite some evidence of the relationship on the international scale (Barrett, submitted for publication), the calls of Adger, Ayers, Paavola, Bird, and the framers of the UNFCCC, are not the instruction in subnational adaptation finance distribution within Malawi. (b) Government model Patronage models offer some explanation for adaptation finance distribution within Malawi, but the finding runs contrary to theory. Districts voting for the incumbent president receive disproportionately less within the capacity development models: within three binary models (models 4–6), the probability of receiving an activity falls from as little as 16.27%, to as much as 27.32%; one significant activity count model indicates a fall from a mean value corresponding to 3.40 adaptation finance activities to 1.33 for districts providing the highest presidential support. Theories of ethno-regionally based patronage suggest districts are rewarded for political support through resource distribution (Azam, 2001; Posner, 2005). However, the restructuring of Malawi politics during 2000–10 offers some explanation. The years leading up to the re-election of incumbent president Bingu wa Mutharika of the Democratic Progressive Party (DPP) in the 2009 witnessed the transformation toward more populist policies (Ferree & Horowitz, 2010; Power, 2010). The decline of ethno-regionalism and the ascent of populism within Malawi suggest the need for a new theoretical basis in which to understand subnational resource allocation dynamics. Mutharika, from the southern district of Thyolo, is from the relatively small Lomwe ethnic grouping. Yet according to the Malawi Electoral Commission, Mutharika, and consequentially the DPP, secured 52% of central presidential votes, which is dominated by the Nyanja ethnic group. Further, in the southern district of Nsanje, neither Mutharika nor his predecessor Muluzu, both of whom are southern, polled more than 33% for their respective parties in the 1999 and 2004 elections. In addition, the climate vulnerability finding suggests another explanation for the weakness of the patronage model. Muthurika and Muluzu have southern constituencies: these are typically the most socio-economically deprived and most climate vulnerable regions within Malawi. If fund allocation is toward wealthy and less climate vulnerable regions, this eventuality is partially conditioning political factors if both presidents are from vulnerable regions. (c) Donor utility Donor utility offers a persuasive explanation for subnational allocation. ODA activity within a district is a consistent positive explanatory factor across the models. The probability of a district receiving a climate-oriented activity rises as little as 63% and as much as 84% (model 1–3) from one ODA activity to the maximum of fifteen; for capacity development activities, the probability rises approximately 65% across models 4–6 from one ODA activity to fifteen. This relationship holds with count models: 12 ODA activities correspond to an average of 1.922 (model 7), 2.060 (model 8) and 2.749 (model 9) climateoriented activities; 12 ODA activities result in an average of 7.044 (model 10), 4.575 (model 11), and 2.749 (model 12) capacity development activities to the district. The findings correspond with other allocation literatures in that the:

(a) minimization of transaction costs determines distribution; and (b) presence of formal aid networks determines the direction of activities regardless of intervention objectives. Therefore, donor utility has a clear role in determining the direction and flow of adaptation finance activities. The proportion of households owning televisions drives allocation, but only for climate-oriented adaptation finance activities. The probability of a district accessing an activity within a given year rises as little as 40% and as much as 59% (model 1–3) from the mean to the maximum value; similarly, activity counts rise from 0.230 to 0.961 (model 7), from 0.246 to 0.830 (model 8) and from 0.225 to 0.966 (model 9) from mean to 7/10ths of maximum value. Counts increase as much as 2.49 activities in districts with the highest engagement with agriculture, but show insignificance in marginal effects models. Therefore, either: (a) governments are directing adaptation finance activities toward politically engaged districts; or (b) the variable is an indicator of income, suggesting multiplier effects, means of lobbying, or good absorptive capacity are driving allocation decisions. The latter corresponds with earlier findings on climate vulnerability: likely recipient districts are those with ample capacity to assimilate and operationalize funds in a productive way. Finally, districts with significant agricultural engagement receive more climate-oriented activities. The probability of a district receiving an activity rises between 14.47% (model 1) and 24% (model 3) from lowest to highest values; likewise, count models show districts receiving 1.406 (model 7), 0.713 (model 8) and 0.611 (model 9) more activities from lowest to highest values. One would expect given the nature of adaptation finance objectives that agriculture would drive allocation. Activities focus on agricultural management, irrigation, and public works programs in areas of high agricultural activity. This suggests that climate-oriented adaptation finance in particular may be fundamentally different from other forms of development finance (Tables 6 and 7). (d) Robustness checks Models are applied with fixed effects to control for stable characteristics and omitted variables within districts (see Tables 8 and 9). If the explanatory factors of adaptation finance distribution show considerable variation across districts, but have little variation over time within districts, then this will create larger standard errors as districts will serve as their own controls. The first point to note is that climate-oriented models dropped 99 observations. Fixed effects models are observing within-district variability and many districts did not retain enough variability for inclusion. Nevertheless, of those with adequate variation we observe the following: The same broad findings are evident from the physical and socio-economic aspects of climate vulnerability. High physical and low socio-economic climate vulnerability drives adaptation finance distribution. Indeed, life expectancy becomes stronger with fixed effects. The same negative relationship is present for government support and is more consistent across these models. The proportion of the district engaged in agriculture drops out in all but model 19, but this remains positive. ODA weakens somewhat in the negative binomial models of climate-oriented activities, but remains a consistent explanatory factor; and television ownership has a similar strong positive relationship within climate-oriented models (13–15 and 19–21) and no relationship to capacity development models (16–18 and 22–24).

SUBNATIONAL CLIMATE JUSTICE? ADAPTATION FINANCE DISTRIBUTION AND CLIMATE VULNERABILITY

137

Table 6. Adaptation finance to districts 2000–2010. Dependent variable equals 1 if activity is made within year, 0 if otherwise; P-values are in parentheses; Models 1–6 use Logit analysis; standard errors are adjusted for 26 district clusters Climate oriented Model 1 CV: EIW

Model 2

Capacity development Model 3

0.792* (0.106)

CV: Quart_Inf_Mort2

Model 4

CV: Quart_Inf_Mort4

Model 6

2.564*** (0.000) 0.486 (0.270) 0.541 (0.168)

0.334 (0.505) 2.203** (0.031) 0.213 (0.738)

CV: Quart_Inf_Mort3

Model 5

0.646** (0.050)

0.254 (0.425) 2.843*** (0.002) 0.254*** (0.000) 0.607*** (0.000) 4.612*** (5.911)

0.034 (0.914) 2.000** (0.035) 0.282*** (0.000) 0.495*** (0.000) 3.666*** (0.000)

0.433* (0.066) 0.075 (0.802) 1.741** (0.052) 0.297*** (0.000) 0.646*** (0.000) 4.003*** (0.000)

R2

16.94%

19.94%

16.99%

22.81%

22.20%

21.81%

N

286

286

286

286

286

286

CV: Life_expect Vote_President Agri_engage ODA Tele_owner Constant

1.010*** (0.002) 0.177 (0.834) 0.537*** (0.000) 0.166 (0.569) 0.578 (0.155)

0.589** (0.023) 0.071 (0.948) 0.547*** (0.000) 0.036 (0.839) 0.934 (0.203)

0.170 (0.141) 0.844*** (0.003) 0.498 (0.512) 0.572*** (0.000) 0.133 (0.461) 0.293 (0.405)

*

p 6 0.10. ** p 6 0.05. *** p 6 0.01.

Table 7. Adaptation finance to districts 2000–2010. Dependent variable of the number of district activities within a given year; P-values in parentheses; Standard errors adjusted for 26 clusters; Models 7–12 use negative binomial regression analysis Climate oriented Model 7 CV: EIW

Model 8

**

Model 10

CV: Quart_Inf_Mort 4

Model 12

0.444 (0.066) 1.073** (0.021) 0.422 (0.169) 0.064 (0.809)

0.597 (0.145) 2.608*** (0.016) 0.204 (0.750)

CV: Quart_Inf_Mort 3

Model 11

*

1.047 (0.024)

CV: Quart_Inf_Mort 2

CV: Life_expect

0.540** (0.029) 0.233 (0.619) 0.114*** (0.016) 0.104 (0.528) 0.666** (0.045)

0.287 (0.208) 0.138 (0.824) 0.118*** (0.009) 0.064 (0.406) 0.555** (0.201)

0.096 (0.250) 0.438 (0.119) 0.162 (0.770) 0.140*** (0.002) 0.076 (0.518) 0.855*** (0.014)

0.123 (0.657) 3.265*** (0.006) 0.251*** (0.000) 0.450** (0.043) 4.525*** (0.000)

0.123 (0.666) 2.122*** (0.017) 0.265*** (0.000) 0.286* (0.087) 3.474*** (0.000)

0.239 (0.352) 0.113 (0.679) 2.211** (0.053) 0.259*** (0.000) 0.458** (0.021) 3.646*** (0.000)

R2

–

–

–

–

–

–

N

286

286

286

286

286

286

Vote_President Agri_engage ODA Tele_owner Constant

*

Capacity development Model 9

p 6 0.10. ** p 6 0.05. *** p 6 0.01.

138

WORLD DEVELOPMENT

Table 8. Adaptation finance distribution to districts 2000–2010. Dependent variable equals 1 if activity is made within year, 0 if otherwise; P-values are in parentheses; Models 13–18 use conditional fixed effects Logit regression analysis Climate oriented Model 13 CV: EIW

Model 14

CV: Quart_Inf_Mort 4

Tele_owner Constant

0.593 (0.164) 3.341 (0.119) 0.193* (0.072) 2.112*** (0.004) –

Model 17

Model 18

2.129*** (0.002)

CV: Life_expect

ODA

Model 16

1.246* (0.073) 0.729 (0.563) 13.506 (0.993)

CV: Quart_Inf_Mort 3

Agri_engage

Model 15

1.742 (0.117)

CV: Quart_Inf_Mort 2

Vote_President

Capacity development

0.101 (0.819) 2.814 (0.180) 0.209** (0.055) 2.143*** (0.011) –

3.627*** (0.000) 0.392 (0.548) 1.666** (0.048) 7.459*** (0.001) 1.269*** (0.008) 2.724 (0.231) 0.167 (0.124) 1.492*** (0.016) –

1.437*** (0.000) 0.034 (0.976) 0.402*** (0.000) 0.431 (0.268) –

0.614* (0.079) 0.732 (0.555) 0.500*** (0.000) 0.508 (0.279) –

2.689*** (0.002) 1.910*** (0.000) 0.511 (0.670) 0.383*** (0.000) 0.524 (0.151) –

R2

–

–

–

–

–

–

N

187

187

187

286

286

286

*

p 6 0.10. p 6 0.05. *** p 6 0.01. **

Table 9. Adaptation finance distribution to districts 2000–2010. Dependent variable represents the number of activities made within a given year; P-values in parentheses; Models 19–24 use conditional fixed effects negative binomial regression analysis Climate oriented Model 19 CV: EIW

Model 20

Model 22

CV: Quart_Inf_Mort 4

Model 23

Model 24

0.699*** (0.004) 0.845* (0.064) 1.191 (0.282) 0.116 (0.904)

CV: Quart_Inf_Mort 3

1.106*** (0.000) 0.438 (0.135) 0.288 (0.378)

0.533* (0.092) 2.990* (0.070) 0.183*** (0.019) 1.649*** (0.002) 4.716*** (0.000)

0.237 (0.475) 2.507 (0.119) 0.118*** (0.018) 1.543*** (0.006) 3.589*** (0.000)

3.762*** (0.000) 0.879*** (0.011) 2.479 (0.116) 0.207*** (0.005) 1.354*** (0.005) 3.548*** (0.000)

R2

–

–

–

–

–

–

N

187

187

187

286

286

286

CV: Life_expect Vote_President Agri_engage ODA Tele_owner Constant

p 6 0.10. p 6 0.05. *** p 6 0.01. **

Model 21

1.237* (0.105)

CV: Quart_Inf_Mort 2

*

Capacity development

0.738*** (0.000) 0.522 (0.335) 0.125*** (0.000) 0.016 (0.883) 0.816*** (0.009)

0.394** (0.020) 0.506 (0.337) 0.090*** (0.003) 0.100 (0.367) 0.496 (0.149)

0.404** (0.036) 0.670*** (0.000) 0.893 (0.113) 0.306*** (0.000) 0.044 (0.686) 0.463* (0.105)

SUBNATIONAL CLIMATE JUSTICE? ADAPTATION FINANCE DISTRIBUTION AND CLIMATE VULNERABILITY

6. CONCLUSION This article investigates subnational adaptation finance distribution and its relation to climate vulnerability, and is the first empirical climate justice study to operationalize adaptation finance within a state. It is the second of three analyses which frame climate justice as a multi-scalar process. The final investigation into local level effectiveness found adaptation finance successful in reducing climate risk at the point of implementation. However, adaptation finance must be present in vulnerable locations to have a positive effect. This article further substantiates that disproportionately less adaptation finance arrives in the areas of highest need. What emerges from this investigation is a picture of the “distribution landscape” built on accessibility, cost minimization, and economic functionality. Districts require pre-existing infrastructure to demonstrate an ability to manage and use funds productively. Conversely, past allocation research suggests a greater, or certainly not an inverse, role for government-district political linkages. The onset of a programmatic political platform during the sample years possibly indicates Malawi as something of an aberration. New subnational geo-coding efforts underway in several other countries provide comparators to test this finding. The article has several theoretical implications. First, researchers often discuss the need to comprehend adaptation finance distribution and climate vulnerability. Developing a recipient need and government interest model provides a means to understand the empirical dynamics of adaptation finance. Iterations, derivatives, and applications of this model have the capacity to construct a broader debate. Second, the research is part of a broad contribution to growing debates on the local consequences of global actions.

139

The implications of this study should concern scholars, activists, and policymakers alike: funds directed to physically vulnerable districts address droughts and floods as hazards, but tend to bypass communities experiencing the greatest climate risk. Wealthy areas have resources that protect against climate variability and change, but communities at the lowest economic, social, and political positions encounter the most adverse consequences of physical hazards, including falling agricultural yields, temporary displacement, disease outbreak, and asset erosion. The provision of proportionately fewer funds to those with the highest climate risk indicates the most eligible are given the least resources to address the inequalities of climate change. The findings above can inform those advocating for vulnerable groups: directing adaptation finance to states at high climate risk will not lead to progress if funds are given to the relatively wealthy within the state. Without knowledge of how adaptation finance flows down to community level implementation, the assumption is that funding reaches the most vulnerable. Yet the actual dynamics of subnational distribution processes may increase inequality for the poorest and most marginalized. This study questions the decision of the global climate regime to re-define “positive action” from mitigation of emissions to the financing of climate adaptation for the most vulnerable. The analysis of climate justice has scope to improve through operationalizing adaptation finance. At present there is no means to trace funds back through multilateral institutions to determine individual state contributions. Second, more agreement on what constitutes “new and additional” climate finance would facilitate greater transparency in data isolation. Third, future research could track historical emissions of individual states and develop a methodology to compare emissions and finance contributions.

NOTES 1. Hulme et al. (2011) explain how the UNFCCC focus to date has been on “‘building adaptive capacity, rather than compensation”’. However, recent talks in Doha have begun a process of recognizing funds given for loss and damage from the adverse effects of climate change as compensation (Clark, 2012). Other factors that build the case for understanding these funds as compensation are outlined in Barrett (2013a). 2. The term “‘activities”’ is preferred over “‘projects”’ for the following reasons: the geo-referencing methodology designed by Strandow, Findley, Nielson, and Powell (2011) use the term “‘projects”’ for international level funding transfers and identify their various geo-referenced components as subnational “‘locations”’; most importantly, the geo-referenced data provides information on where the specific activity components of these broader projects are being implemented. These are not to be confused with village or group-village level projects or interventions that are at a much finer grain of disaggregation. 3. By definition, ODA addresses poverty alleviation and must conform to the requirements of primarily addressing economic development and being concessional in character (IMF, 2003). Poverty is a driver of climate vulnerability, which offers one explanation for the assimilation of CFfA into the greater body of development aid. However, CFfA is designed to lessen sensitivity and exposure to a physical threat, which is itself determined by a complex political, social and economic process only partially explained by poverty (Ayers & Forsyth, 2009; Wisner et al., 2004).

4. Decision-making is conducted through committees: District Executive Committees exercise executive responsibilities, which are supported by District Development Committees —– composed of political and traditional leaders and heads of sector ministries, Area Development Committees and Village Committees (Local Government Forum, 2012). 5. Narrow CFfA refers to international transfers that make an explicit reference to adaptation in direct association with climate change. The working definition is any financial transfers with the explicit objective of facilitating adaptation or strengthening adaptive capacity to climate change. Conversely, broadly defined CFfA were located by whether these projects sought “‘vulnerability reduction”’ in terms of the direct and indirect effects of climate change. The operational definition of broad CFfA is any financial transfers that contribute to climate vulnerability reduction. Some cases have clear climate vulnerability reducing objectives with reference to aspects of climate resilience, climate-proofing and coping with climate change. For projects that have no direct climate change reference, inclusion is based on whether there is reference to chronic and sudden-onset disasters brought about by drought, desertification, land degradation, floods, hurricanes, temperature increases and sea-level rise, as suggested by the Intergovernmental Panel on Climate Change (IPCC) report (2001). Hence, the necessary and sufficient conditions of inclusion rest on (a) whether any one of these direct and indirect effects of climate change are addressed; and (b) whether the project title or description contains a number of features, characteristics and idiosyncrasies related to vulnerability reduction in relation to climate change (for elaboration please see —- Barrett, submitted for publication).

140

WORLD DEVELOPMENT

6. The complete list of active donors to Malawi between during 2000-– 2010 is as follows: African Development Bank, Australian Oversees Aid Program, Arab Bank for Economic Development in Africa, Canadian government, center for Disease Control and Prevention, Department for International Development, European Union, Food and Agricultural organization for the United Nations, Flemish International Cooperation Agency, German Agency for Technical Cooperation, Icelandic International Development Agency, Irish government, Japanese government, Japan International Cooperation Agency, (KFW) Reconstruction Credit Institute, Kuwait Fund, Norwegian government, OPEU Fund, People’s Republic of China, United Nations Development Program, United Nations Educational, Scientific and Cultural Organization, (UNHCR) United Nations Human Rights Agency, United Nations Industrial Development organization, USAID, World Food Program, World Bank. 7. The CCAPS method relies upon a coding exercise that draws from actual donor project documents collected via Malawi’s Aid Management Platform and through direct contact with aid donors in the country. The coder reads each project document in full and identifies all activities within the project. Each activity is then geo-coded and climate coded simultaneously by two research assistants, with discrepancies reconciled by a senior coder (arbitrator). Thus each project is vetted by three trained coders, with an inter-coder reliability rate of over 84% percent. . .. . . The CCAPS methodology codes for climate relevance of aid activities using a continuous spectrum. The spectrum includes four poles, ranging from Ambiguous Development (which provides the least benefit to adaptation, including maladaptation) to Climate-Oriented (which is explicitly designed to address climate issues). In between these two categories: Capacity Development reflects activities that enhance resilience to climate change, but without that purpose in mind, and General Development reflects activities that enhance human and environmental well-being but are not explicitly driven by or obviously directly relevant to address climate change threats (Weaver et al., 2012, p. 2).

8. Weaver et al. (2012), p. 5. 9. Projects may have no direct climate change reference, but there is an association to climate change chronic and sudden-onset disasters brought about by drought, desertification, land degradation, floods, hurricanes, temperature increases and sea-level rise, as suggested by the IPCC report 2001. 10. Geo-coding is either presently underway, or plans are in motion, in Haiti, Nepal, Laos, Senegal and Uganda. 11. The Environmental Indications and Warnings Project under the Global Climate Change Research Program is sponsored and funded by the Central Intelligence Agency. Any opinions, findings, and conclusions, or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the CIA or the U.S. Government. 12. EIW include socio-economic indicators, but disaggregated time-series data are only available for political atrocities and internal war, neither of which are applicable to the Malawi context. 13. This study first investigated district linkages to president’s ethnicity, language and religion, but none captured any clear patterns; second, the margin of victory uncovered the same broadly negative relationship as with the percentage of vote; third, examining whether a district returned a candidate of the presidents party was not possible due to Bingu Mutharika changing parties while in office. 14. The intercept and regression coefficients are estimated using the maximum likelihood procedure.

REFERENCES Ackerson, C., Adams, E., Arnold, J., Magero, V., Knop-Narbutis, A., & Phillips, P. (2011). Measuring international development assistance for climate change adaptation in Africa (Unpublished Manuscript). Austin: University of Texas. pp. 38 Adger, W. N., Paavola, J., Huq, S., & Mace, M. (2006). Fairness in adaptation to climate change. Cambridge: MIT Press. Anderson, S. (2013). Climate justice and international development: Policy and programming, IIED Policy Brief, 1–4. Arler, F. (2001). Global partnership, climate change and complex equality. Environmental Values, 10(3), 301–329. Ayers, J. (2009). International funding to support urban adaptation to climate change. Environment and Urbanization, 21(1), 225–240. Ayers, J., & Forsyth, T. (2009). Community-based adaptation to climate change: Strengthening resilience through development. Environment, 51(4), 22–31. Ayers, J. M., & Huq, S. (2009). Supporting adaptation to climate change: What role for official development assistance? Development Policy Review, 27(6), 675–692. Azam, J. P. (2001). The redistributive state and conflicts in Africa. Journal of Peace Research, 38(4), 429–444. Barrett, S. (2013a). The necessity of a multi-scalar analysis of climate justice. Progress in Human Geography, 37(2), 215–233. Barrett, S. (2013b). Local level climate justice? Adaptation finance and vulnerability reduction. Global Environmental Change, 23(6), 1819–1829. Barrett, S. (submitted for publication). An international analysis of climate justice: Adaptation finance distribution and climate vulnerability. Global Environmental Politics. Benson, T. (2002). An atlas of social statistics: Malawi. Washington, DC: International Food Policy Research Institute.

Besley, T., & Burgess, R. (2002). The political economy of government responsiveness: Theory and evidence from India. Quarterly Journal of Economics, 117(4), 1415–1451. Beymer-Farris, B. A., & Bassett, T. J. (2012). The REDD menace: Resurgent protectionism in Tanzania’s mangrove forests. Global Environmental Change – Human and Policy Dimensions, 22(2), 332–341. Bird, N., Brown, J., & Schalatek, L. (2011). Design challenges for the green climate fund. Climate finance policy brief. Berlin: Heinrich Boll Stiftung (pp. 8). Blaikie, P., Cannon, T., Davies, I., & Wisner, B. (1994). At risk: Natural hazards, people’s vulnerability, and disasters. London: Routledge. Busby, J., Smith, T., & White, K. (2011). Locating climate insecurity: Where are the most vulnerable places in Africa? CCAPS Policy Brief 3. Austin: Robert Strauss Centre (pp. 12). Busby, J., Smith, T., White, K., & Strange, S. (2013). Climate change and insecurity: Mapping vulnerability in Africa. International Security, 37(2), 132–172. Caney, S. (2006). Cosmopolitan justice, rights and global climate change. Canadian Journal of Law and Jurisprudence, 19(2), 255–275. Caney, S. (2010). Markets, morality and climate change: What, if anything, is wrong with emissions trading? New Political Economy, 15(2), 25. Chandrasekhar, S. (1965). American aid and India’s economic development. London: Praeger. Chidanti-Malunga, J. (2011). Adaptive strategies to climate change in Southern Malawi. Physics and Chemistry of the Earth, 36(14–15), 1043–1046. Ciplet, D., Roberts, T., & Khan, M. (2013). The Politics of International Climate Adaptation Funding: Divisions in the Greenhouse. Global Environmental Politics, 13(1), 49–68.

SUBNATIONAL CLIMATE JUSTICE? ADAPTATION FINANCE DISTRIBUTION AND CLIMATE VULNERABILITY Clark, P. (2012). Doha talks agree to climate compensation, Financial Times, December 8. Clay, D. C., Molla, D., & Habtewold, D. (1999). Food aid targeting in Ethiopia: A study of who needs it and who gets it. Food Policy, 24(4), 391–409. Commonwealth Local Government Forum (2012). Commonwealth local government handbook. London: Publications UK Limited. Davies, G. M., Pollard, L., & Mwenda, M. D. (2010). Perceptions of land degradation, forest restoration and fire management: A case study from Malawi. Land Degradation & Development, 21(6), 546–556. Department of Environment Malawi (2006). National adaptation programme of action, NAPA. Lilongwe: Ministry of Mines, Natural Resources and Environment (pp. 45). Dulal, H., Brodnig, G., Thakur, H. K., & Green-Onoriose, C. (2010). Do the poor have what they need to adapt to climate change? A case study of Nepal. Local Environment, 15(7), 612–635. Eriksen, S., & Lind, J. (2009). Adaptation as a political process: Adjusting to drought and conflict in Kenya’s drylands. Environmental Management, 43(5), 817–835. Fazlul, M., & Nobuo, M. (2008). Impacts of climate change and sea-level rise on cyclonic storm surge floods in Bangladesh. Global Environmental Change, 18(3), 490–500. Ferree, K., & Horowitz, J. (2010). Ties that bind? The rise and decline of ethno-regional partisanship in Malawi, 1994–2009. Democratization, 17(3), 534–563. Forsyth, T., & Sikor, T. (2013). Forests, development and the globalization of justice. The Geographic Journal, 179(2), 114–121. Francken, N., Minten, B., & Swinnen, J. (2009a). The political economy of relief aid allocation: Evidence from Madagascar. LICOS Discussion Paper (pp. 31). Leuven: LICOS Centre for Institutions and Economic Performance. Francken, N., Minten, B., & Swinnen, J. F. M. (2009b). Media, monitoring, and capture of public funds: Evidence from Madagascar. World Development, 37(1), 242–255. Francken, N., Minten, B., & Swinnen, J. F. M. (2012). The political economy of relief aid allocation: Evidence from Madagascar. World Development, 40(3), 486–500. Garg, A. (2011). Pro-equity effects of ancillary benefits of climate change policies: A case study of human health impacts of outdoor air pollution in New Delhi. World Development, 39(6), 1002–1025. Global Climate Change Research Program. (2011a). Freshwater security anomalies: Ten year composite (10-year threshold; January 2002– December 2011). Environmental Indications and Warnings Project. Central Intelligence Agency, US Government. Global Climate Change Research Program. (2011b). Freshwater security anomalies: Five year composite (10-year threshold; January 2007– December 2011). Environmental Indications and Warnings Project. Central Intelligence Agency, US Government. Global Climate Change Research Program (2011c). Freshwater security anomalies: Three year composite (10-year threshold; January 2009– December 2011). Environmental Indications and Warnings Project. Central Intelligence Agency, US Government. Green, R. M. (1977). Intergenerational distributive justice and environmental responsibility. Bioscience, 27(4), 260–265. Griniski, S., Collins, T., Ford, P., Fitzgerald, R., Aldouri, R., VelazquezAngulo, G., et al. (2012). Climate change and environmental injustice in a bi-national context. Applied Geography, 33, 25–35. Hulme, M., O’Neill, S. J., & Dessai, S. (2011). Is weather event attribution necessary for adaptation funding? Science, 334(6057), 764–765. Intergovernmental Panel on Climate Change (2001). Climate change 2001: Synthesis report. Cambridge: Cambridge University Press. International Monetary Fund (2003). External debt statistics: Guide for compliers and users. Washington, DC: International Monetary Fund. Jayne, T. S., Strauss, J., Yamano, T., & Molla, D. (2001). Giving to the poor? Targeting of food aid in rural Ethiopia. World Development, 29(5), 887–910. Jayne, T. S., Yamano, T., Weber, M. T., Tschirley, D., Rui, B. F., & Chapoto, A. (2003). Smallholder income and land distribution in Africa: Implications for poverty reduction strategies. Food Policy, 28(3), 253–275. Lopez-Marrero, T. (2010). An integrative approach to study and promote natural hazards adaptive capacity: A case study of two flood-prone communities in Puerto Rico. Geographical Journal, 176(2), 150–163. Maltais, A. (2008). Global warming and the cosmopolitan political conception of justice. Environmental Politics, 17(4), 592–609.

141

Markandya, A. (2011). Equity and distributional implications of climate change. World Development, 39(6), 1051–1060. Mendelsohn, R., Morrison, W., Schlesinger, M. E., & Andronova, N. G. (2000). Country-specific market impacts of climate change. Climatic Change, 45(3–4), 553–569. Mercer, K. L., Perales, H. R., & Wainwright, J. D. (2012). Climate change and the transgenic adaptation strategy: Smallholder livelihoods, climate justice, and maize landraces in Mexico. Global Environmental Change–Human and Policy Dimensions, 22(2), 495–504. Mijoni, P. L., & Izadkhah, Y. O. (2009). Management of floods in Malawi: Case study of the Lower Shire River Valley. Disaster Prevention and Management, 18(5), 490–503. Morris, S. S., & Wodon, Q. (2003). The allocation of natural disaster relief funds: Hurricane Mitch in Honduras. World Development, 31(7), 1279–1289. Moser, C. (2008). Poverty reduction, patronage, or vote buying? The allocation of public goods and the 2001 election in Madagascar. Economic Development and Cultural Change, 57(1), 137–162. Mustafa, D. (1998). Structural causes of vulnerability to flood hazard in Pakistan. Economic Geography, 74(3), 289–305. Paavola, J., & Adger, W. N. (2006). Fair adaptation to climate change. Ecological Economics, 56(4), 594–609. Page, E. (2011). Cosmopolitanism, climate change and greenhouse gas emissions. International Theory, 3(1), 37–69. Paxson, C., & Schady, N. R. (2002). The allocation and impact of social funds: Spending on school infrastructure in Peru. World Bank Economic Review, 16(2), 297–319. Peratsakis, C., Powell, J., Findley, M., Baker, J., & Weaver, C. (2012). Climate coded and geocoded activity-level data from the government of Malawi’s aid management platform. Washington, DC: AidData and the Robert S. Strauss Center for International Security and Law. Phiri, I. M. G., & Saka, A. R. (2008). The impact of changing environmental conditions on vulnerable communities in the Shire Valley, Southern Malawi. Future of Drylands, 545–559. Posner, D. N. (2005). Institutions and ethnic politics in Africa. Cambridge: Cambridge University Press. Power, J. (2010). Political culture and nationalism in Malawi: Building Kwacha. Woodbridge: University of Rochester Press. Rawls, J. (1972). A theory of justice. Oxford: Oxford University Press. Reinikka, R., & Svensson, J. (2004). Local capture: Evidence from a central government transfer program in Uganda. Quarterly Journal of Economics, 119(2), 679–705. Roberts, T., & Peratsakis, C. (2011). Measuring ODA for climate change: Comparing 19 categorizations of DFID’s 2,226 FY2008-2009 projects (Unpublished Manuscript, pp. 46). Rhode Island: Brown University. Roberts, T., & Parks, B. (2007). A climate of injustice. Cambridge: MIT Press. Stadelmann, M., Persson, A., Ratajczak-Juszko, I., & Michaelowa, A. (2013). Equity and cost-effectiveness of multilateral adaptation finance – Are they friends or foe? International Environmental Agreements: Politics, Law and Economics, 1, 1–20. Strandow, D., Findley, M., Nielson, D., & Powell, J. (2011). The UCDP and AidData codebook on georeferencing aid Version 1.1. UCDP Paper (pp. 30). Uppsala: Uppsala University. Stringer, L. C., Dyer, J. C., Reed, M. S., Dougill, A. J., Twyman, C., & Mkwambisi, D. (2009). Adaptations to climate change, drought and desertification: Local insights to enhance policy in southern Africa. Environmental Science and Policy, 12(7), 748–765. Suiseeya, K., & Caplow, S. (2013). In pursuit of procedural justice: Lessons from an analysis of 56 forest carbon project designs. Global Environmental Change, 23(5), 968–979. Takasaki, Y. (2011). Do local elites capture natural disaster reconstruction funds. Journal of Development Studies, 10(1), 18. United Nations Framework Convention on Climate Change. (1992). Rio: United Nations. Weaver, C., Baker, J., & Peratsakis, C. (2012). Tracking climate adaptation aid: Methodology. Climate Change and African Political Stability Research Brief No. 5. Austin: University of Texas (pp. 7). Weaver, C., & Peratsakis, C. (2011). Can better tracking of adaptation aid reduce climate change vulnerabilities on the ground? Climate Change and African Political Stability Research Brief 2. Austin: University of Texas (pp. 7). Wheeler, D. (2011). Quantifying vulnerability to climate change: Implications for adaptation assistance. Washington, DC: Center for Global Development (pp. 53).

142

WORLD DEVELOPMENT

Wisner, B., Cannon, T., Davies, I., & Blaikie, P. (2004). At risk: Natural hazards, people’s vulnerability and disasters (2nd ed.). London: Routledge. World Bank (2013). World development indicators CO2 emissions data. Washington, DC: The World Bank Group. Zhang, G. (2004). The determinants of foreign aid allocation across China: The case of World Bank loans. Asian Survey, 44(5), 691–710.

APPENDIX A. DEPENDENT VARIABLE DETAILS Climate oriented: Coding isolates 123 entries during 2000– 10. Data are aggregated at the district level by year using GIS shapefiles, involving 286 district-year units, and ranging between 0 and 4 activities. Capacity development: Coding isolates 846 entries of capacity development adaptation finance during 2000–10. Adaptation finance data are aggregated at the district level by year using GIS shapefiles, and provides 286 district-year units, ranging between 0 and 34 activities.

APPENDIX B. EXPECTED VARIABLE MOVEMENTS The expectation is for b1 to be positive in this arrangement since higher climate vulnerability raises demand and supply for CFfA distribution within districts suffering climate-related challenges. There is an anticipation that b2 will be positive, due to the necessity of governments to reward districts who are political supporters with resource distribution. This is most prominent in African states where client-based politics is well theorized, due to the ethno-regional political party formations that are understood to structure elite behavior. b3 can also be forecast as holding a positive value, as the larger the general development aid network within the district, the more amenable the environment for further donor activity. b4 should hold a positive value given that greater absorption capacity makes the district a suitable location to gain a greater yield in terms of climate resilience. However, if b4 is positive, this will be conditional upon b1 holding minus values as the socio-economic indicators within Cvit will have to be inversely related to CFfA distribution. b5 can be expected to be positive due to the fact that CFfA projects are often designed for agricultural areas. In turn, this suggests that climate-oriented CFfA may have a stronger positive effect than capacity development CFfA.

Available online at www.sciencedirect.com

ScienceDirect