WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS K E N YA’ S VA N I S H I N G H E R D S Richard Damania, Sébastien Desbureaux, Pasquale Lucio Scandizzo, Mehdi Mikou, Deepali Gohil, Mohammed Said When Good Conservation Becomes Good Economics Kenya’s Vanishing Herds Produced with support from Table of Contents List of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii At a crossroads: Safari tourism under threat in Kenya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Following the tracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix The trade-off between road construction and wildlife protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi Building smart infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii The promise of conservancies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii The race between conservancies and construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii Good conservation is good economics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv 1. Vanishing Herds: Wildlife Dynamics and Drivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 A declining tourism asset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Declining natural assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2. Weighing the Impacts: Generating Scenarios and Simulating Trade-offs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Developing a regional ESAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 The CGE model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Rural poverty and tourism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Concluding comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3. Wildlife-Friendly Roads: Fable or Fact? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 New approaches to enhancing road access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Conventional approaches of increasing road access . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Counting the costs of business as usual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 A greener scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Fine-tuning the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 4. The Way Forward and Next Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Smart infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Realizing economic opportunities through conservancies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Appendix A. Conservancies—An Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 The environmental promise of conservancies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 The economic significance of conservancies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Going further . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 Appendix B. Road Extension and Wildlife Loss between 1980 and 2010: A Difference-in-Differences Approach . . . 48 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50   iii List of Figures Figure ES.1: Kenya has witnessed a dramatic collapse in wildlife since the 1980s . . . . . . . . . . . . . . . . . . . . vii Figure ES.2: Wildlife is now found in fragmented habitats and has vanished across vast areas in the North . . . . viii Figure ES.3: Roads lead to changes in land use, impacting wildlife most severely within a distance of 20 kilometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Figure ES.4: Production frontier for GDP and loss of wildlife . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi Figure ES.5: Factoring in wildlife constraints significantly reduces the impact of new roads on wildlife . . . . . . . xii Figure ES.6: A map of Kenya’s conservancies and parks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii Figure 1.1: Kenya has lost 68 percent of its wildlife in recent decades . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Figure 1.2: Wildlife trends in the 19 rangeland counties between 1977 and 2016 (percent) . . . . . . . . . . . . . . . 2 Figure 1.3: Kenya’s wildlife populations have shrunk dramatically, becoming fragmented, and almost vanishing in some counties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Figure 1.4: Human populations have been expanding in areas with wildlife and around parks . . . . . . . . . . . . . 4 Figure 1.5: Kenya’s road network has increased by 50 percent in the last 40 years . . . . . . . . . . . . . . . . . . . 5 Figure 1.6: Roads lead to changes in land use, impacting wildlife most severely within a distance of 20 kilometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Figure B.1.1: Declining wildlife . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Figure B.1.2: Declining wildlife . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Figure 2.1: Mapping the CGE regions for Kenya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Figure 2.2: Forward multipliers for productive sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13 Figure 2.3: Backward multipliers for productive sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14 Figure 2.4: CGE model simulation for base-year production activities . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Figure 2.5: Multipliers as a function of investment size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17 Figure 2.6: Value added variation in regard to wildlife reduction and tourism elasticity . . . . . . . . . . . . . . . . 21 Figure 2.7: Relationships between wildlife reduction and value added, and value added growth . . . . . . . . . . . 21 Figure B2.1: Impact on environmental deterioration against GDP growth . . . . . . . . . . . . . . . . . . . . . . . . 22 Figure 2.8: Income elasticities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Figure 2.9: Income elasticities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Figure 3.1: Existing all-weather roads and tracks in Kenya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Figure 3.2: The costs of increasing Kenya’s RAI under the “business as usual” scenario . . . . . . . . . . . . . . . 30 Figure 3.3: Building new roads to increase Kenya’s RAI, starting with the densely populated western counties . . 30 Figure 3.4: Wildlife loss is constant for the first 1.5 million people connected to the road network, with losses sharply increasing thereafter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .31 Figure 3.5: The costs of increasing Kenya’s RAI under the two scenarios . . . . . . . . . . . . . . . . . . . . . . . . 32 Figure 3.6: Factoring in wildlife constraints significantly reduces the impact of new roads on wildlife . . . . . . . . 33 Figure 3.7: Mapping elephant and wildebeest routes in Kenya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Figure 4.1: Map of parks and conservancies in Kenya (2018) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Figure 4.2: Wildlife generally increased in the older conservancies and decreased in areas where conservancies were established after 1995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Figure A.1: The rapid growth of conservancies in Kenya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Figure A.2: Tourism income earned by conservancies (Ksh, 2017) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Figure A.3: Proportion of conservancy income sources in 2017 (Ksh, millions) . . . . . . . . . . . . . . . . . . . . . 45 iv  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds Figure B.1:  Kenya’s wildlife populations have shrunk dramatically since the 1980s, becoming fragmented, and almost vanishing in some counties, such as in West Pokot, Baringo, Turkana, Machakos, Kwale, and Mandera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Figure B.2: Distance to roads and wildlife loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 List of Tables Table ES.1: Regional impacts of investing in conservancies and construction . . . . . . . . . . . . . . . . . . . . . . xiv Table 1.1: Quantifying the impact on wildlife of construction of a road (wildlife biomass) . . . . . . . . . . . . . . . . . 8 Table 2.1: CGE investment impact multipliers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Table 2.2: Impacts of investment in conservancies (50% of current value = $142 million) . . . . . . . . . . . . . . . .17 Table 2.3: Impacts of investment in construction ($142 million) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Table 2.4: Impact on regional value added of an increase in infrastructure (+ 10%) and reduction in wildlife in the South (percent from baseline) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Table 2.5: Impact on regional income distribution of an increase in infrastructure and reduction in wildlife in the South (percent from baseline) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Table 2.6: Impact on regional value added of an increase in investment in road construction (+10%) and greater reduction in wildlife in the South (percent from baseline) . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Table 2.7: Impact on income distribution of an increase in infrastructure and a greater reduction in wildlife in the South (percent from baseline) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Table 2.8: Doubling the investment in conservancies: impact on value added . . . . . . . . . . . . . . . . . . . . . 23 Table 2.9: Doubling the investment in conservancies: impact on income distribution . . . . . . . . . . . . . . . . . 24 Table 2.10: Doubling investment in conservancies and wildlife conservation: impact on value added . . . . . . . . 24 Table 2.11: Doubling investment in conservancies and wildlife conservation: impact on income distribution . . . . . 24 Table 2.12: Doubling investment and capital productivity in conservancies and wildlife conservation: impact on value added . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Table 2.13: Doubling investment and capital productivity in conservancies and wildlife conservation: impact on income distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Table 4.1: GDP multipliers for investments (in million USD) in conservancies . . . . . . . . . . . . . . . . . . . . . . . 36 Table A.1: Typology of Kenyan conservancies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Table A.2: An overview of tourism facilities in Kenya’s conservancies . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Table B.1: Main model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Table B.2: Parsimonous model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 List of Boxes Box 1.1:  Building a Statistical Model to Quantify the Direct Impact of Roads on Wildlife . . . . . . . . . . . . . . . . . 8 Box 1.2:  Population Density and Wildlife Loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Box 2.1:  Key Assumptions of the CGE Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14 Box 2.2:  Trade-offs between Economic Growth and Environmental Impact . . . . . . . . . . . . . . . . . . . . . . 22 Box 4.1:  Smart Infrastructure and Spatial Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Box A.1:  Some Key Figures on the Economics of Conservancies in Kenya . . . . . . . . . . . . . . . . . . . . . . . 43 Box A.2:  Types of Benefit-Sharing Arrangements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Table of Contents  v List of Acronyms CGE Computable General Equilibrium DRSRS Department of Resource Surveys and Remote Sensing ESAM Environmentally-extended SAM Ksh Kenya shilling KWCA Kenya Wildlife Conservancies Association KWS Kenya Wildlife Service RAI Rural Access Index SAM Social Accounting Matrix SDGs Sustainable Development Goals TLU Tropical Livestock Unit vi  Executive Summary At a crossroads: Safari tourism international tourist arrives on a package tour that may include a safari, a visit to the beach, or both. It is safari under threat in Kenya tourism, however, that generates the most employment It is no exaggeration to state that Kenya’s wildlife has and economic activity across the country. A recent study done much to shape the image and development for- by Sanghi et al. (2017) found that despite a diversifying tunes of the country. At independence, the country was economy, wildlife-based safari tourism is deeply inte- reliant on agricultural exports for its foreign exchange grated into Kenya’s economic fabric in complex ways that revenue and was exposed to the vagaries of commodity stimulate much employment in rural areas. Official statis- price cycles. The vast and varied endowment of wildlife tics of the sector’s contribution to the economy tend to catalyzed a new industry—nature-based tourism—that neglect the full panoply of backward and forward linkages provided an opportunity to diversify and boost export and their dynamic effects on poverty and rural growth. revenues while playing to the country’s natural compara- tive advantage. But the wildlife that has lured travelers to Kenya by the planeload is in dramatic decline (Figure ES.1). In the past Today tourism is among Kenya’s top sources of foreign three decades, the country has lost more than half of its exchange, dominates the service sector, and contrib- wildlife (ungulate) biomass according to data from the utes significantly to employment, especially in rural areas Directorate of Resources, Surveys and Remote Sensing where economic opportunities are limited. The typical (DRSRS). FIGURE ES.1: Kenya has witnessed a dramatic collapse in wildlife since the 1980s Wildlife Loss in Kenya, 1980–2000 1,800 1,600 Total wildlife biomass (in 1,000 kg) 1,400 1,200 1,000 800 1980 1985 1990 1995 2000 Decade Source: Authors based on DRSRS data.   vii FIGURE ES.2: Wildlife is now found in fragmented habitats and has vanished across vast areas in the North Source: Authors. Data from DRSRS and Ogutu et al. (2016). Wild herds that once roamed freely across the borders ecosystem and determines whether a species will sur- of Kenya and Tanzania have shrunk dramatically in num- vive.1 The process of dispersing from a natal territory is bers and vanished completely from much of the North essential to avoid inbreeding and it strongly influences (Figure ES.2). Once connected habitats have been sev- individual fitness. ered, with herds trapped into shrinking areas, jeopar- dizing the long-term sustainability of many isolated and As a result, wildlife depends as much on adjacent land unconnected populations. for continued viability as it does on the protected areas. Pressures around the parks are affecting wildlife within Perhaps most troubling is that recent monitoring of the parks. The way in which land outside of protected wildlife populations suggests that long-term declines of areas is utilized and managed will become a crucial many of the charismatic species that attract tourists—­ determinant of the industry’s future. Expanding tour- lions, elephants, giraffes, impalas, and others—are ism to these areas remains among the most successful occurring at the same rates within the country’s national approaches that have been piloted. However, the feasi- parks as outside of these protected areas (Ogutu et al. bility of this approach depends upon economic incen- 2016). Parks in Kenya were established in areas in which tives and the opportunity costs of land. large aggregations of animals were observed typically during the dry seasons, but in their haste to establish these protected areas, policy makers neglected the 1  Dispersal is a fundamental behavioral and ecological process. The distance migratory needs of wildlife, especially of the ungulate that individual animals disperse, and the number of dispersers, can be primary herds. Dispersal is a fundamental biological process determinants of where and whether species persist. Dispersal fundamentally influences spatial population dynamics, including meta-population and meta- that influences the distribution of biodiversity in every community processes. viii  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds This report uses a variety of approaches to investigate uses, and examines the extent to which land conversion the economic consequences of this decline. State-of- leads to the extirpation of wildlife and the loss of tourism the-art spatial econometric methods are used to identify incomes. the causal drivers of the loss and quantify the impacts on wildlife. A Computable General Equilibrium (CGE) model The analysis presented in Chapter 1 finds that roads are is used to estimate the economic consequences of wild- typically accompanied by a change in land use pattern life loss and compare these consequences to alternative from natural habitats to farms and settlements. On aver- development pathways. Finally, spatial algorithms are age, the extent of conversion is especially stark up within developed to show how losses can be avoided and how a corridor of about 20 kilometers from the road. Thereaf- to create win-win solutions that maximize economic gains. ter, the conversion of natural habitats into cropland slowly decreases and is almost negligible for settlements. An obvious consequence of this change in land use is the Following the tracks almost complete collapse of wildlife in areas around the Reasons for the decline in Kenya’s wildlife have been roads (Figure ES.3). The statistical model developed for widely documented, and they entail an interconnected this report indicates that roads built over the last four suite of pressures typically linked to habitat conversion— decades have caused an 80 percent decrease in wild- factors such as population growth, the expansion of life within a 20-kilometer radius. There are also predict- arable agriculture, fencing, poaching, and intrusive infra- able effects on migratory corridors, which have almost structure. This report identifies with greater precision the all been diminished and degraded to varying degrees drivers of land conversion from natural habitats to other (Ojwang et al. 2017). FIGURE ES.3: Roads lead to changes in land use, impacting wildlife most severely within a distance of 20 kilometers Natural habitat converted in croplands and distance to roads 1.2 Average area converted in cropland per cell (ha) 1 .8 .6 .4 .2 0 20 40 60 80 100 Distance to nearest road (km) Source: Authors based on ESA land use data and Michelin roads data. Note: TLU stands for Tropical Livestock Units. Executive Summary  ix FIGURE ES.3: Continued Natural habitat converted in settlement and distance to roads .3 Average area converted in settlement per cell (ha) .2 .1 0 0 20 40 60 80 100 Distance to nearest road (km) Total loss of wildlife between 1980–2000s and distance from roads in 2000s 350 Total loss of wildlife biomass (in 1,000 TLU) 300 250 200 150 0 20 40 60 80 100 Distance to nearest road in km Source: Authors based on ESA land use data and Michelin roads data. Note: TLU stands for Tropical Livestock Units. x  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds The report then provides an assessment of the eco- wildlife tourism loss are much more severe in the North nomic consequences of this loss. Clearly, if the eco- where there are limited development opportunities. nomic benefits brought about by habitat conversion outweigh the losses, it is arguable that the extirpation This relationship is summarized in Figure ES.4, which of wildlife is a necessary, if regrettable, price to pay for shows the production possibility frontier of the Kenyan development. But if the loss of actual and potential tour- economy. If a road brings losses of wildlife that are below ism income exceeds the benefits from land conversion, a threshold (around 30 percent), it confers a net eco- greater care and caution would be warranted in both the nomic benefit and an increase in GDP. But losses that are placement of intrusive developments and the extent of much larger induce a net loss in GDP. Put simply, good land conversion. conservation has become good economics for Kenya. FIGURE ES.4: Production frontier for GDP and loss of wildlife The trade-off between road 60,000 construction and wildlife protection 59,500 Value added ($, millions) 59,000 To explore this issue in a rigorous manner, this report 58,500 employs a computable general equilibrium (CGE) model 58,000 that divides the economy into two regions—North and 57,500 South. The model tracks the contribution and linkages 57,000 56,500 between various economic activities and provides an 56,000 indication of the economic consequences of alternative 55,500 development strategies (Chapter 2). 55,000 0% 20% 40% 60% 80% 100% Reduction of wildlife The projections indicate that the two regions have differ- Source: Authors. ent economic structures. In general, land-based activi- ties, manufacturing, trade, and transport are the sectors that create the largest gains (production multiplier One implication of this finding is that the induced con- effects) in the economy. Production multipliers, though version of habitats has come at a high cost to much of low in both regions of the country, tend to be compara- the country. A second implication is that since the poten- tively larger in the more developed areas of the South. tial and often hidden benefits of habitats are significant, development opportunities exist to harness the dual Has the loss of wildlife generated economic gains benefits of both conservation and development. Finally, commensurate to the economic loss? A road through these results also suggest that if the consequences rural areas brings multiple benefits through the expan- of construction were managed and controlled better sion of agriculture, access to markets, and myriad eco- so that habitat conversion was prevented and wild- nomic opportunities that such market integration brings. life losses avoided, it might be possible to simultane- Accordingly, the CGE model finds that if the consequent ously obtain the benefits of infrastructure development loss of wildlife is relatively modest and below around as well as those brought by tourism. This would likely 30  percent (or alternatively, if the elasticity of tourism entail significant and different policy interventions. The with respect to wildlife is small), there is limited loss of available data suggest that the declaration of protected tourism and there is a net gain to the conversion of land. area status or conservancy status may slow, though not prevent, the rate of land conversion for agriculture and If, on the other hand, the loss of wildlife is much larger, settlements. The report explores two sets of solutions to there is a decline in overall regional GDP. With the maximize the benefits of infrastructure and of tourism: 80 percent loss of wildlife experienced in Kenya within a road network that pays attention to the externalities 20 kilometers of a road, parts of the country would no that it generates, and a policy that expands the role of doubt fall into the latter category. In general, impacts of conservancies. Executive Summary  xi Building smart infrastructure In sum, deploying smarter, greener approaches to infrastructure also makes economic sense. Achieving The key to avoiding the economic costs identified in this equilibrium will call for more sophisticated plan- this report is to find ways to maximize the benefits from ning approaches that recognize both the benefits as infrastructure and minimize the economic losses. About well as the adverse impacts for both the economy and 30 percent of Kenya’s rural population is currently con- wildlife. nected to the national all-season road network. Increas- ing the country’s Rural Access Index (RAI) will be key to achieving the goals set in Kenya’s Vision 2030 and under the Global Sustainable Development Goals. The promise of conservancies Conservancies can play an important role in diversifying Using state-of-the art algorithms, this report finds that the tourism product and securing critical habitats while the judicious location of roads can connect much of generating economic activity. There are currently more the country to centers of economic activity while avoid- than 166 conservancies spread across Kenya’s 28 coun- ing potential losses of wildlife. This is because much ties (Figure ES.6). They cover an area larger than the coun- of Kenya’s densely populated western counties require try’s national parks, are home to more than 22 percent of rural roads, but they are also areas with low levels of Kenya’s ungulate wildlife biomass, and have some of the wildlife and tourism potential. Figure ES.5 illustrates highest densities of wildlife in the country. In fact, 18 out one such example and shows that with sophisticated of the 20 zones with the highest density of wildlife are planning approaches, equivalent “connections” can be in conservancies and not parks. Conservancies create made with much more limited losses to wildlife (com- buffers around parks and maintain connectivity between pare the green and orange lines) and at roughly the several ecosystems. In essence, conservancies are key same cost. to the resilience of wildlife. FIGURE ES.5: Factoring in wildlife constraints significantly reduces the impact of new roads on wildlife 200,000 175,000 150,000 Impacted wildlife (kg) 125,000 100,000 75,000 50,000 25,000 0 0 500,000 1,000,000 1,500,000 2,000,000 2,500,000 3,000,000 Population with new access to all-weather roads Current situation More wildlife friendly roads Source: Authors. xii  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds FIGURE ES.6: A map of Kenya’s conservancies and parks Source: Authors. Tourism remains an important revenue stream for con- remain among the few ways in which communities can servancies, accounting for an average of 83 percent boost and diversify income sources. of commercial revenue. In many of the conservancies, tourism facilities were established to create an exclu- The race between conservancies sive game viewing experience as an alternative to the mass tourism strategies in neighboring national parks and construction and reserves. The game lodges in the conservancies It is instructive to determine the economic benefits of alter- account for about 16 percent of the total bed-nights native investment strategies in contexts when there are spent in Kenyan game lodges, suggesting considerable limited resources available for expansion. Using the CGE scope for expansion. In remote areas, conservancies model, Table ES.1 provides an indication of the benefits Executive Summary  xiii TABLE ES.1: Regional impacts of investing unlikely to do so in the future when pressures expand and in conservancies and construction competition for land, water, and other natural resources intensifies. This suggests an urgent need for a careful Invest in South Invest in North reassessment of pressures, policies, and future prospects. Construction multiplier 1.51 0.88 0.1 1.53 Conservancy multiplier 3.02 0.22 1.75 4.41 Wildlife in Kenya, especially in the North of the country, Source: Authors. represents a lucrative economic asset whose contribution has been underestimated and potential unrealized. Con- that accrue from investing in a road in each region of the verting habitats and dissecting wildlife migration corridors country, which is compared to investing in conservancies. diminishes populations, tourism appeal, and the earning potential of natural assets in ways that are often irreparable The table shows the payoffs from a road-building invest- and irreversible. Given the significant and long-term impli- ment in the South. Each dollar invested in the South gen- cations of such decisions, a rigorous economic assess- erates on average a GDP increase of $1.51 in the South ment is necessary to guide choices. The CGE assessment and $0.8 in the North. An equivalent investment in the indicates that every dollar invested in conservation and North has a similar multiplier effect, so that every dollar wildlife tourism could generate benefits that range from $3 invested in the North has a payback of $1.53, but this to $20, with returns that increase with the level of invest- time with a much smaller spillover to the South (.01).2 ment (Chapter 4). Such increasing returns reflect the eco- The North has historically lagged in economic terms. logical importance of connected natural habitats that are The investment in tourism offers high payoffs with the more productive in terms of the ecosystem services that promise of igniting economic activity in ways that also they provide and are also more resilient to droughts and contribute to environmental sustainability in an arid area other weather extremes. With the right infrastructure and with geographic constraints. Realizing this promise will the enabling environment to further develop the conser- require enabling policies that provide access to conser- vancy sector, there are significant opportunities to enhance vancies and share the benefits with the population. growth through the conservation of wildlife assets. A similar investment in conservancies generates signifi- The evidence presented in this report suggests that cantly higher multipliers—almost twice as high. This is a there are wide opportunities to stop the dramatic col- consequence of the important role that wildlife plays in lapse of wildlife populations and that investing in the the tourism value chain, with multiple direct and indirect tourism sector yields significant benefits which are espe- connections to employment-generating activities in sec- cially pro-poor. The most pressing need is for planners to tors that themselves have high multipliers, such as trans- incorporate the tools developed in this report and else- portation and lodging. where in order to consider the long-term implications of irreversible decisions and harness the full potential that the country’s natural endowment offers. Good conservation is good economics A 70 percent decline in wildlife, within thirty years, is a REFERENCES sobering statistic. As Kenya’s population grows, its infra- Ogutu, Joseph O., et al. (2016). “Extreme wildlife declines and concurrent increase in livestock numbers in Kenya: What structure needs expand, and climate change makes rain- are the causes?” PloS one 11.9. fall more erratic; the pressures on wildlife and natural Ojwang, Gordon O., et al. (2017). Wildlife Migratory Corridors habitats will intensify in regions that are already under and Dispersal Areas: Kenya Rangelands and Coastal Ter- environmental stress and will spread to other parts of the restrial Ecosystems. country. The journey along the current policy path has Sanghi, Apurva, Richard Damania, Farah Manji, and Maria failed to halt the degradation of natural habitats, and it is Paulina Mogollon. (2017). Standing out from the herd: An economic assessment of tourism in Kenya (English). Wash- 2  The magnitude of these multipliers is similar to global estimates. ington, D.C., World Bank Group. xiv  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds CHAPTER 1 VANISHING HERDS: WILDLIFE DYNAMICS AND DRIVERS Wildlife, the principal asset of Kenya’s tourism indus- FIGURE 1.1: Kenya has lost 68 percent of its wildlife in try, is in rapid decline. To some this may be a neces- recent decades sary, if regrettable, price to pay for development and an 2,500,000 expanding economy. For others, declining wildlife num- bers are associated with costs, including lower tourism revenues, which could have been avoided with a less 2,000,000 intrusive development trajectory. The aim of this report Population estimates is to explore these issues using a suite of rigorous eco- nomic modeling approaches. The report combines sta- 1,500,000 tistical approaches to determine what has happened, with macroeconomic modeling to answer counterfactual questions regarding what might have happened with 1,000,000 alternative policies. The overall analysis suggests that the economic impacts 500,000 of natural capital erosion have been significant, and they 1977 1986 1995 2004 2013 have received less policy attention than seems war- Year ranted since these issues are viewed as environmental Source: DRSRS, Ogutu et al. (2016). problems that drain public funds, rather than an eco- nomic loss. The focus of this chapter is on tracking the use changes, land fragmentation, infrastructure develop- changes and dynamics of Kenya’s key tourism asset—its ment (Sala et al. 2000; EC 2001), poaching (WWF 2017), wildlife. Subsequent chapters explore the economic con- climate change (Wiens 2016), and other factors are among sequences of this loss and then turn to policy options. At the long litany of reasons given for this rapid decline the outset several caveats must be noted. First, due to (Dybas 2009; Daskin and Pringle 2018; WWF 2017). insufficient data this report is narrowly focused on herds of (charismatic) mammals and thus ignores other spe- In Kenya, wildlife has declined precipitously across the cies, as well as ecosystem productivity. In addition, the country, and for certain species, this decline has been cat- investigation is restricted to the measurable and pecuni- astrophic. Within three decades, Kenya has lost 68  per- ary benefits generated by conservation through tourism. cent of its wildlife (Figure 1.1). The declines were particularly Consideration of the wider benefits (such as watersheds) extreme with a wide cross-section of species that includes conferred by ecosystems would suggest that the value ungulates and predators.3 As a consequence, in 2018, of Kenya's wildlands are much higher than is suggested in this report. 3  To be precise the declines were: warthog (–87.7 percent), waterbuck (–87.8 percent), Grevy’s zebra (86.3 percent), impala (–84.1 percent), Coke hartebeest (84 percent), topi (–82.1 percent), oryx (–78.4 percent), eland A declining tourism asset (–77.7 percent), Thomson’s gazelle (–75 percent), and lesser kudu (–72.4 percent). The declines were also severe for Grant’s gazelle (–69.6 percent), gerenuk (68.6 percent), giraffe (–66.8 percent), and wildebeest (–64.2 percent). Globally, there is mounting evidence of catastrophic In comparison ostrich (–43.4 percent), elephant (–42.3 percent) buffalo (–36.9 percent), and Burchell’s zebra (29.5 percent) experienced moderate declines in the number and range of wildlife populations declines. Similar downward trends were exhibited by the big cats and other (Ceballos et al. 2017). Rapid human population growth, land carnivores as their populations have also declined rapidly (Virani et al. 2011).   1 FIGURE 1.2: Wildlife trends in the 19 rangeland counties between 1977 and 2016 (percent) Taita Taveta West Pokot Tana River Machakos Samburu Mandera Marsabit Turkana Baringo Laikipia Garissa Kajiado Kwale Narok Lamu Isiolo Wajir Kitui Kilifi 0 –7 –20 –41 –40 –50 –56 –60 –60 –60 –66 –71 –70 –76 –79 –80 –85 –86 –87 –84 –87 –90 –92 –99 –100 Source: Authors based on Ogutu et al. (2016). Kenya was ranked 5th in Africa in terms of the number of the Tropical Livestock Unit (TLU) where 250 kilograms is threatened species within its country (IUCN 2018). equivalent to 1 TLU).4 The losses have occurred across the entire country, with The changes in the status of wildlife has been striking. some variation over the 19 counties. The highest decline In the 1980s, around 53 percent of the (5 × 5 kilome- was observed in West Pokot, which has experienced ter) grids in the 19 counties were occupied by wildlife. a total collapse, with 99 percent of its wildlife lost. The By the 2000s, this had fallen to 31 percent of grid cells. smallest decline of wildlife was observed in Laikipia, Figure 1.3 provides a summary of these data and shows which experienced a 7 percent decrease in wildlife bio- extirpation over large areas of the country. The distribu- mass (Figure 1.2). The three other major tourist-­dependent tion map of the 2000s indicates that the wild herds that counties of Narok, Kajiado, and Taita Taveta showed once roamed freely across the country have shrunk dra- varying trends: Narok, despite its high dependence on matically in numbers and distribution, and have vanished wildlife-based tourism, has lost about 70 percent of its in counties such as West Pokot, Turkana, Baringo, Kilifi, wildlife; in Kajiado, the decline stands at 60 percent; and Lamu, Machakos, and Tana River. Once connected habi- Taita Taveta registered a moderate decrease of about tats have been severed and isolated, with herds trapped 40 percent. This suggests that the presence of buoyant into shrinking areas, which affects their long-term sus- wildlife-based tourism in a county may not be sufficient to tainability (Said et al. 2016). counter the forces behind the decline in wildlife. This also implies that there is a need for a deeper understanding of Perhaps more troubling is that recent monitoring efforts of the drivers of wildlife loss to counter the problem. key species suggest that the long-term decline of many including of the charismatic species that attract tourists—­ Data from the Department of Resource Surveys and lions, elephants, giraffes, and impalas—are occurring at Remote Sensing (DRSRS) provide a more precise indica- tion of trends and drivers of change. DRSRS has con- 4  Eighteen species were used in the analysis: buffalo (Syncerus caffer), Burchell’s zebra (Equus burchelli), Coke hartebeest (Alcelaphus buselaphus), ducted aerial surveys of wildlife in the rangelands of eland (Taurotragus oryx), elephant (Loxodonta africana), gerenuk (Litocranius Kenya since 1977, offering a uniquely rich database of walleri), giraffe (Giraffa cemelopardalis), Grant’s gazelle (Gazella granti), Grevy’s zebra (Equus grevyi), impala (Aepyceros melampus), lesser kudu (Tragelaphus wildlife population trends at a fine spatial scale. Within imbermbis), oryx (Oryx gazelle beisa), ostrich (Struthio camelus), Thomson’s each grid, wildlife populations for 18 common species gazelle (Gazella thomsoni), topi (Damaliscus lunatus korrigum), warthog (Pharcoerus africanus), waterbuck (Kobus ellipsiprymnus), and wildebeest are measured in terms of their biomass (calculated using (Connochaetes taurinus). 2  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds FIGURE 1.3: Kenya’s wildlife populations have shrunk dramatically, becoming fragmented, and almost vanishing in some counties Source: Authors based on DRSRS data. comparable rates within and outside protected areas include population growth (Kenya’s population has grown (Scholte 2011). This is consistent with a growing body of more than sixfold since 1961), the expansion of arable evidence in the conservation literature, which finds that agriculture, fencing, poaching, and intrusive infrastruc- the creation of protected areas does not necessarily mean ture (Said et al. 2016) (Figure 1.4). This report expands that habitats and species are effectively protected (Andam upon this literature by providing quantitative estimates et al. 2008), and that stricter rules on land use do not nec- of some of the drivers of the loss of wildlife—something essarily translate into less degradation (Ferraro et al. 2013). that, to our knowledge, has been done for the first time. Parks in Kenya were established in areas where large PAVING THE WAY aggregations of animals were observed, typically dur- ing the dry seasons. However, in the process of estab- Roads are a formidable engine for growth and poverty lishing these protected areas, policy makers neglected reduction. They connect people to jobs, schools, markets, the migratory needs of wildlife, especially the ungulate and hospitals. In rural areas, they improve market access herds. Dispersal is a fundamental biological process for farmers, allowing them to sell their products at higher that influences the distribution of biodiversity in every prices, thus raising incomes. Roads boost the develop- ecosystem and determines whether a species will sur- ment of commercial agriculture, aiding in the transition vive. Among other things, the process of dispersing from subsistence to market economies. New roads also from a natal territory is essential to avoid inbreeding and connect people to the rest of society, which creates a strongly influences individual fitness. As a result, wildlife shared existence and builds a larger identity. For these depends as much on adjacent land as it does on the pro- reasons and more, increasing the rural road network is tected areas for continued viability. Between 60–80 per- central to the Sustainable Development Goals (SDGs). cent of the wildlife in Kenya is found outside protected Specifically, SDG indicator 9.1.1 encourages policy mak- areas (Grunblatt et al. 1996; Western et al. 2009). ers to increase the share of the rural population who live within 2 kilometers of an all-season road that is motorable all year round by the prevailing means of rural transport. In Declining natural assets the relatively dry environment of Kenya, paved as well as improved roads can be considered as all-season roads.5 Reasons for the decline of Kenya’s wildlife have been widely documented and involve an interconnected suite 5  In countries with more wet conditions, it is often only paved roads that are of pressures typically linked to habitat conversion. These considered to be all-weather roads. Vanishing Herds: Wildlife Dynamics and Drivers  3 FIGURE 1.4: Human populations have been expanding in areas with wildlife and around parks Source: Authors based on Michelin and WorldPop data. Indeed transport networks such as the railways have overlaying data from WorldPop, which provides estimates played a key part in the development of Kenya. of population density at a precise spatial scale of about 1 kilometer, with the 2017 road network, an index of acces- Kenya’s road network has grown considerably over the sibility to roads can be derived. This index is termed the last decades. Michelin maps of East Africa, dating back Rural Access Index (RAI) (Stevens et al. 2015). The results to 1978, were digitized to investigate the expansion and show that only about 28 percent of the rural population in effects of roads. In 1978, there were around 7,000 kilo- Kenya lives within 2 kilometers from a road—an RAI that meters of paved and improved roads (Figure 1.5), and is comparable to most developing countries. Significant the entire North of the country only featured improved opportunities therefore exist to connect Kenya’s rural gravel roads at this time. In the subsequent 40 years, population to the main network, and continued invest- Kenya’s road network had increased by 50 percent to ments in road infrastructure may serve as a key lever to cover around 11,000 kilometers of improved and paved reducie poverty and promote inclusive growth. The chal- roads as of 2017. The network of roads has become lenge for the country is to achieve this in ways that do not denser in the South but has also been extended in the diminish the economic value of its natural assets. North to connect the major urban centers in the region, an example being the recent paving of roads leading to CONNECTIONS THAT DISCONNECT Marsabit and Turkana Counties. While roads bring important benefits to people, there is a Despite the significant and successful extension of Kenya’s growing body of evidence suggesting that they may also road network, the majority of the country’s rural popula- generate significant environmental impacts, especially in tion continues to live more than 2 kilometers from an all- areas with rich biodiversity. A large and rapidly expand- season road. Population growth has far outpaced efforts ing literature has documented the impact of roads on for- to connect the country’s centers of agglomeration. By est cover across countries as diverse as Brazil (Laurance 4  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds FIGURE 1.5: Kenya’s road network has increased by 50 percent in the last 40 years et al. 2014); the Democratic Republic of Congo (DRC) DRC, for example, a significant impact on deforestation (Damania et al. 2018); India (Asher et al. 2018); Indone- is seen up to 2 kilometers from roads, and in Western sia, Tanzania (Arcus Foundation 2018); and at a global Tanzania, the impact is seen even f ­arther—deforestation scale (Arcus Foundation 2018). Studies consistently find even increased 20 to 30 kilometers away from the newly that the extension of roads into forested areas catalyzes built Ilagala–Rukoma–­Kashagulu Road (Asher et al. 2018). deforestation or forest degradation, though the magni- In general, the scale of habitat loss is determined by the tude of impact differs considerably across countries.6 This incentives unleashed to expand cropland into natural occurs not only through the direct clearing of vegetation habitats and the capacity to regulate these. There are to open up the road, but mainly from the indirect threats likely other effects, such as the spread of invasive spe- brought by people settling around the new roads, who cies, that are ignored in this report. now benefit from easier access to markets, which leads to the conversion of natural habitats into croplands. In the QUANTIFYING THE CAUSAL IMPACTS In Kenya too, statistical analyses indicate that roads are 6  Asher et al. (2018) find no effect of local roads on deforestation in India, but a large impact of national roads on deforestation. a key part of this dynamic and have predictable effects Vanishing Herds: Wildlife Dynamics and Drivers  5 on wildlife and on migratory corridors. Because land use road (control). Simultaneity bias may be a significant threat change is the primary driver of biodiversity loss, it is no when studying the impact of roads on wildlife since wild- surprise that the rate of wildlife loss in Kenya between life distribution and road placement are jointly determined. the 1980s and the end of the 2000s was significantly For example, new roads may be targeted for regions faster in areas in close proximity to roads. Almost all with expanding agricultural activity and land use, imply- wildlife corridors have been affected by land conversion, ing that these roads may be a response to activities that though the extent varies (Ojwang et al. 2017). are already causing forest cover reduction. Difference- in-differences models combined with fixed effects are an Nonparametric statistical models known as LOWESS effective method to overcome this challenge (Asher et al. regressions were used to investigate the causal links, if any, 2018). The approach presents what may be the first causal between natural habitat loss in Kenya and the distance to estimates for Kenya by exploiting spatial location and roads. The approach uses state-of-the-art statistical mod- timing—comparing rates of biodiversity loss of a cell that ­ els to identify the causal impact of roads and isolate these remains at a large distance from a road to one that was from the confounding effects (Ali et al. 2015). The Euclidean once a large distance away but has been brought close distance between each grid cell and the nearest paved or to the road. Time, cell, and other fixed effects control for improved road was calculated for each decade from the other factors to address omitted variable bias and other 1980s to the 2000s. These distances were then catego- problems. The full model is presented in Box 1.1. rized into different bins depending on whether a cell was less than 5, 10, 15, 20, or 50 kilometers from a road. The results, illustrated in Figure 1.6, show that the closer a grid cell is to a road, the faster the conversion of natural Difference-in-differences models were used to estimate habitat to cropland and settlements, which consequently the change in wildlife biomass inside cells that, over time, has an impact on wildlife. Results from the statistical came into closer proximity to a road (treatment) com- model consistently suggest that cells located within a pared to those cells that remained farther away from a 20-kilometer distance to a road are associated with a FIGURE 1.6: Roads lead to changes in land use, impacting wildlife most severely within a distance of 20 kilometers Natural habitat converted in croplands and distance to roads 1.2 Average area converted in cropland per cell (ha) 1.0 0.8 0.6 0.4 0.2 0 20 40 60 80 100 Distance to nearest road (km) Source: Authors based on DRSRS, ESA, and Michelin data. 6  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds FIGURE 1.6: Continued Natural habitat converted in settlement and distance to roads .3 Average area converted in settlement per cell (ha) .2 .1 0 0 20 40 60 80 100 Distance to nearest road (km) Total loss of wildlife between 1980–2000s and distance from roads in 2000s 350 Total loss of wildlife biomass (in 1,000 TLU) 300 250 200 150 0 20 40 60 80 100 Distance to nearest road in km Source: Authors based on DRSRS, ESA, and Michelin data. Vanishing Herds: Wildlife Dynamics and Drivers  7 BOX 1.1:  Building a Statistical Model to Quantify the Direct Impact of Roads on Wildlife Data compiled by the Department of Resource Surveys and Remote Sensing (DRSRS) using aerial surveys in the rangelands of Kenya since 1977 cover 19 rangeland counties. The approach used for estimating impacts follows best practices. Kenya was divided into a grid of 10 × 10 kilometers, and for each cell, wildlife in the 1980s were identified in each pixel. Changes in wildlife measured in TLU were then determined for the 1990s and 2000s. In addition, the average distance between each grid cell and the nearest road between 1978 and 2010 was also determined. A “difference-in-differences” specification is used to determine the impact of roads on wildlife loss. The model exploits the expansion of the road network in Kenya in the 1980s–90s. Between 1978 and 1992, the average distance of a cell’s centroid from a road went from 55 km to 44 km (10% decrease). Cells that were originally (1980s) far from a road (50–100 km) are kept in the analysis. Among these cells, the model looks at how the loss of wildlife differed between cells that became closer to a road (treatment groups, 5 km, 10 km, 15 km, 20 km, and 50 km) and cells that remained far from a road (control group, >50 km). Roads here include both paved and improved roads. Formally, the model is: Wildlifei,t = β Cell Close from Roadi,t + γ Posti,t + ω Cell Close from Road ∗ Posti,t + η Protected Areai,t + i + µt + ∈i,t Where: —Wildlifei,t: Total biomass of wildlife in cell i during decade t (t = 1980, 1990, 2000). —Cell Close from Road (Treatment): Whether the cell has become 5, 10, 15, or 20 km closer to a road during the period. —Post: Dummy variable for post 1980s decade (i.e., once most cells became close to a road). Roads of the 1980s = roads observed in 1978; roads in the 1990s = roads observed in 1992; roads in the 2000s = roads — observed in 2003. —Additional controls: Dummy variable for the presence of a protected area in the cell, province × decade fixed effect. —Clustered standard errors. Weights based on the area of each cell. The methodology is further detailed in Appendix B. TABLE 1.1: Quantifying the impact on wildlife of construction of a road (wildlife biomass) Variables Less than 5 km Less than 10 km Less than 15 km Less than 20 km Less than 50 km Treated × post –217.369* –185.138** –134.558* –114.494* –65.554 (121.325) (85.765) (79.109) (65.091) (44.057) Post –358.628*** –389.410*** –345.288*** –326.846*** –530.919*** (101.365) (100.153) (110.082) (105.581) (143.272) Observations 2,586 2,730 2,868 3,027 4,029 Number of cells 862 910 956 1,009 1,343 Treatment Road becomes <5 km Road becomes <10 km Road becomes <15 km Road becomes <20 km Road becomes <50 km Control Road 50 to 100 km Road 50 to 100 km Road 50 to 100 km Road 50 to 100 km Road 50 to 100 km from cell from cell from cell from cell from cell Note: * = p<0.05, ** = p<0.01, *** = p<0.001. 8  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds BOX 1.2:  Population Density and Wildlife Loss The question of how population growth impacts natural habitats and wildlife is at the heart of policy debates, at least since Hardin’s (1968) seminal assessment of the Tragedy of the Commons. While the tragedy can be avoided (Boserup 1965; Ostrom 1990), numerous studies have empirically established a correlation between population growth and environment degradation. Population growth was found to be associated with losses of both natural habitats such as forests and savan- nahs (Cropper and Griffiths 1994; Jha and Bawa 2006) and wildlife (Du Toit and Cumming 1999). Notably, these correlations were observed in East Africa and Kenya (Du Toit and Cumming 1999; Ogutu et al. 2011; Ogutu et al. 2016; Veldhuis et al. 2019), where demographic growth remains high today (2.4 percent annually in the Seregenti-Mara region (Veldhuis et al. 2019)). Figures B.1.1 and B.1.2 illustrate the correlations that exist between population density and wildlife in Kenya using the data compiled for this report. Figure B.1.1 shows the correlation between ungulate wildlife biomass between 2000 and 2010, and total human population using 5 km × 5 km gridcells. It shows a negative correlation between ungulate wildlife density in the 2000s and human population. Indeed, in the 5 km × 5 km grid cells where human population is under 500 inhabitants, ungu- late wildlife density is estimated at about 40 Tropical Livestock Units (TLU). However, in grid cells where human population reaches 4,000 inhabitants, almost no wildlife is found any more. Figure B.1.2 illustrates a dynamic. It plots the relationship between the rate of ungulate wildlife loss between the 1980s and the 2000s, and human population in the same grid. Once again, it shows that the rate of wildlife loss is positively correlated with human population: the more a grid cell is populated, the larger is the wildlife loss. The results in this section illustrate that for any given population density, the construction of a road will hasten and intensify the decline in wildlife. A long list of economic literature highlights that the development of infrastructure—particularly roads, is a leading determi- nant of where population growth happens: people follow infrastructure since it offers economic opportunities. Therefore, the current choices made regarding infrastructure construction will have long-lasting impacts on the demography of the country, and consequently consequences on future wildlife trends. As demonstrated in the rest of this report, large room exists to build infrastructures in key economic areas and protect wildlife at the same time. FIGURE B.1.1  Declining wildlife FIGURE B.1.2  Declining wildlife Wildlife density and population Wildlife loss and total population 50 .92 Wildlife density in the 2000s (in TLU) Wildlife loss 1980s–2000s (in %) 40 .90 30 .88 20 .86 10 .84 0 .82 0 1,000 2,000 3,000 4,000 0 1,000 2,000 3,000 4,000 Total population in 2010 Total population Vanishing Herds: Wildlife Dynamics and Drivers  9 significant decrease in wildlife following construction of Ceballos, G., P. R. Ehrlich, and R. Dirzo. (2017). “Biological Anni- the road, and the closer a cell is to a road, the larger the hilation via the Ongoing Sixth Mass Extinction Signaled by impact. Vertebrate Population Losses and Declines.” Proceedings of the National Academy of Sciences. 114 (30). Cropper, M., and Griffiths, C. (1994). The interaction of popula- To be specific, the results in Table 1.1 suggest that a cell tion growth and environmental quality. The American Eco- that was once 50 kilometers away from a road and that nomic Review, 84(2), 250–254. has been brought to within 5 kilometers of a road will Damania, R., et al. (2018). “The Road to Growth: Measuring have lost an additional 217 Tropical Livestock Units the Tradeoffs between Economic Growth and Ecological (TLUs) (or 54,250 kilograms of wildlife biomass) over a Destruction.” World Development 101. decade compared to cells that remained 50 kilometers Daskin, J. H., and R. M. Pringle. (2018). “Warfare and Wildlife from a road. Given that the average wildlife biomass in a Declines in Africa Protected Areas.” Nature 553, 328–332. cell between 1980 and 2009 was 266 TLU, the impact of Du Toit, J. T., and Cumming, D. H. (1999). Functional significance roads has been significant. The estimates suggest that of ungulate diversity in African savannas and the ecologi- in the first 5 kilometers from a road, wildlife loss is the cal implications of the spread of pastoralism. Biodiversity & most severe, at 80 percent (217/266). Wildlife loss falls Conservation, 8(12), 1643–1661. to 69 percent at a distance of 5 to 10 kilometers, 50 per- Dybas, C. L. (2009). “Infectious Diseases Subdue Serengeti cent at a distance of 10 to 15 kilometers, and 40 percent Lions.” BioScience 59, 8–13. EC. (2001).“Biodiversity in Development.” Biodiversity Brief 8. at a 20-kilometer distance. Hence, even after 20 kilo- Ferraro, Paul J., et al. (2013). “More Strictly Protected Areas are meters from a road, the impact remains ecologically sig- not Necessarily More Protective: Evidence from Bolivia, nificant though much smaller. Overall, and on average, a Costa Rica, Indonesia, and Thailand.” Environmental road results in a decline of 76 percent of wildlife biomass Research Letters 8:2. within a 20-kilometer radius. Grunblatt, J., M. Y. Said, and P. Wargute. (1996). National Range- lands Report: Summary of Population Estimates of Wildlife Having identified the causal impact of roads on wildlife, and Livestock (1977–1994). DRSRS, Ministry of Planning it is necessary to determine if the resulting gains have and National Development, Nairobi, Kenya. outweighed the forgone losses of tourism revenue. An Hardin, G. (1968). The tragedy of the commons. Science, economic model of Kenya is used to answer this ques- 162(3859), 1243–1248. tion in the next chapter, followed by a discussion of win- IUCN. (2018). “The IUCN Red List of Threatened Species.” Ver- win solutions to these problems. sion 2018-2: http://www.iucnredlist.org. Downloaded on 17 March 2019. Laurance, W. F., et al. (2014). “A Global Strategy for Road Build- ing.” Nature 513. REFERENCES Ogutu, J. O., Hans-P. Piepho, M. Y. Said, G. O. Ojwang, L. W. Ali, Rubaba, et al. (2015). Highways to success or byways to Njino, S. C. Kifugo, and P. W. Wargute. (2016). “Extreme waste: Estimating the economic benefits of roads in Africa. Wildlife Declines and Concurrent Increase in Livestock Washington DC: The World Bank. Numbers in Kenya: What are the Causes?” PLoS ONE 11 (9). Andam, Kwaw S., et al. (2008). “Measuring the Effectiveness Ogutu, J. O., Owen-Smith, N., Piepho, H. P., and Said, M. Y. of Protected Area Networks in Reducing Deforestation.” (2011). Continuing wildlife population declines and range Proceedings of the National Academy of Sciences 105.42. contraction in the Mara region of Kenya during 1977–2009. Arcus Foundation. (2018). State of the Apes: Infrastructure Journal of Zoology, 285(2), 99–109. Development and Ape Conservation, (Cambridge: Cam- Ojwang, Gordon O., et al. (2017). Wildlife Migratory Corridors bridge University Press—Chapter 3). and Dispersal Areas: Kenya Rangelands and Coastal Ter- Asher, S., T. Garg, and P. M. Novosad. (2018). The Ecological restrial Ecosystems. Impact of Transportation Infrastructure. Washington DC: Ostrom, E. (1990). Governing the commons: The evolution of World Bank. institutions for collective action. Cambridge University Press. Boserup, Ester. (1965) The conditions of agricultural growth: Said, M. Y., J. O. Ogutu, S. C. Kifugo, O. Makui, R. S. Reid, and The economics of agrarian change under population J. de Leeuw. (2016). “Effects of Extreme Land Fragmenta- pressure. tion on Wildlife and Livestock Population Abundance and Distribution.” Journal for Nature Conservation 34: 151–164. 10  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds Sala, O. E., et al. (2000). “Global Biodiversity Scenarios for the Scavenging Raptors in and around the Masai Mara Ecosys- Year 2100.” Science 287 (5459), 1770–1774. tem, Kenya.” Biological Conservation. 144, 746–752. Scholte, P. (2011).“Towards Understanding Large Mammal Pop- Western, D., S. Russell, and I. Cuthill. (2009). “The Status of ulation Declines in Africa’s Protected Areas: A West-Central Wildlife in Protected Areas Compared to Non-Protected African Perspective.” Tropical Conservation Science. Areas of Kenya.” PLoS One 4. Stevens, Forrest R., et al. (2015). “Disaggregating Census Wiens, J. J. (2016). “Climate-related local extinctions are already Data for Population Mapping Using Random Forests with widespread among plant and animal species.” PLoS Biol- Remotely-Sensed and Ancillary Data.” PLoS ONE 10:2. ogy, 14, e2001104. Veldhuis et al. (2019). Science 363, 1424–1428. WWF. (2017). Poaching and Illegal Wildlife Trade. WWF, Virani, M., C. Kendall, P. Njoroge, and S. Thomsett. (2011). Switzerland. “Major Declines in the Abundance of Vultures and Other Vanishing Herds: Wildlife Dynamics and Drivers  11 CHAPTER 2 WEIGHING THE IMPACTS: GENERATING SCENARIOS AND SIMULATING TRADE-OFFS The aim of this chapter is to examine the economic con- FIGURE 2.1: Mapping the CGE regions for Kenya sequences of the trade-offs that confront policy makers. While roads are necessary for development, they bring economic costs through the loss of tourism income. The magnitude of gains and losses involved is unknown, rendering policy choices difficult and questionable. This chapter attempts to provide answers to these far-­ reaching and difficult issues. To do so, it relies on two analytical tools: a regional social accounting matrix (SAM) extended to a set of environ- mental accounts (environmentally-extended SAM or ESAM), and a computable general equilibrium (CGE) model. Consistent with the characteristics of a new gen- eration of applied economic models (Perali and Scan- dizzo, 2018), SAM and ESAM provide a way of linking Kenya’s national accounts to investment scenarios and policy changes in order to estimate impacts on growth, jobs, incomes, exports, and other key economic and social indicators, as well as environmental variables. While the ESAM provides the data for the exercise, the CGE is the engine (the model) that simulates impacts. Source: Method developed by the World Bank. It remains one of the most rigorous quantitative meth- ods for generating economically consistent scenarios to Appendix B, Scandizzo and Ferrarese (2015), and Sanghi evaluate the impact of economic and policy shocks. The et al. (2017). model used for this report is an extension of an earlier model that was used to assess the economic impacts of The ESAM describes an economy with strong dualistic tourism in Kenya (see Sanghi et al. 2017). features, where the South is vastly more developed than the North, and where inter-sector linkages tend to rein- force a pattern of concentration of economic activities Developing a regional ESAM in the more advanced South. While the Southern value The ESAM estimated for Kenya divides the country into chains have depth, especially in agriculture and other two parts—the South and the North—as shown in Fig- land- and food-related activities, the region is still highly ure  2.1. It comprises 30 sectors for each region, and dependent on imports in the manufacturing sectors (Fig- several environmental sectors and factors, as well as ures 2. 2 and 2.3). household types and institutional accounts (govern- ment, capital formation, and rest of the world). Details To gain a better understanding of the structure of the of the methodology and the estimates are contained in economy, it is instructive to examine the multipliers in 12  FIGURE 2.2: Forward multipliers for productive sectors 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 Agriculture Livestock Forestry Fishing Wildlife Parks Conservancies Water biodiversity Mining Food, beverage, and tobacco All other manufacturing Distribution water Utilities Construction Trade Accommodation and restaurant Lodge Transport Information and communication Financial and insurance activities Real estate Business and administrative Other services Public administration and defense Health and social work Education Park tourism Beach tourism Cultural tourism Business tourism South North Source: Elaboration of the Kenya SAM. the model. In a CGE context, “forward” multipliers mea- depth and interconnectedness. However, backward mul- sure the degree to which a sector participates in an over- tipliers in the South are without exception higher than in all expansion (or contraction) of the economy, i.e., the the North. increase required in the supply of one sector to meet a uniform increase of demand, spread over all sectors. Conversely, “backward” multipliers measure the degree to which a sector is capable of stimulating other sectors The CGE model through an increase in the demand for inputs. A back- While the SAM multipliers may give a first approxima- ward multiplier thus indicates the amount of output gen- tion of the indirect effects of investment and other policy erated in an economy due to an exogenous increase in changes, they do not take into account the more com- the demand in a given sector. plex secondary impacts on employment and prices. These effects are likely to be important when explor- The multipliers tend to be comparatively larger in the ing economic changes of significance, such as a large more developed areas of the South, where forward mul- investment or a policy shift. Box 2.1 provides an overview tipliers are much larger in sectors such as agriculture, of the key assumptions of the CGE model. trade, transport, manufacturing, and financial services. Differences are smaller in natural resource–based sec- Figure 2.4 illustrates the ability of the model to track the tors and ecotourism, reflecting the comparative advan- Kenyan economy. Model calibration achieves almost a tage of the North. Backward multipliers are low in both perfect fit, except for the category of “all other manufac- regions, suggesting that value chains still lack overall turing activities,” which is a residual. Weighing the Impacts: Generating Scenarios and Simulating Trade-Offs   13 FIGURE 2.3: Backward multipliers for productive sectors 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 Agriculture Livestock Forestry Fishing Wildlife Parks Conservancies Water biodiversity Mining Food, beverage, and tobacco All other manufacturing Distribution water Utilities Construction Trade Accommodation and restaurant Lodge Transport Information and communication Financial and insurance activities Real estate Business and administrative Other services Public administration and defense Health and social work Education Park tourism Beach tourism Cultural tourism Business tourism South North Source: Elaboration of the Kenya SAM. BOX 2.1:  Key Assumptions of the CGE Model The CGE is built on the assumption of a “small economy,” in the sense that the country cannot influence international prices of imported and exported goods. Each sector produces a composite commodity that can either be exported or produced for the domestic market. Each producer is assumed to maximize profits by producing one commodity, with labor, capital, land, and ecosystem services as primary inputs, according to a constant elasticity of substitution (CES) production function. The demand for intermediate inputs assumes fixed input-output coefficients, and the demand for primary factors is given by the first order condition for profit maximization using value added prices. Production is either for the domestic market in each region or for trade/exports with the other region or the international market according to a Constant Elasticity of Transfor- mation (CET) function. Producers are assumed to maximize revenue from sales subject to the CET function. Export supply represents the first order condition and is a function of the elasticity of transformation and the relative export price with respect to domestic price. The allocation of imports and domestic production is determined according to the hypothesis that domestic and internationally traded goods are imperfect substitutes that are combined in a composite good according to a constant elasticity technology. Aggregate domestic demand is divided into four components for both regions: consumption, intermediate demand, gov- ernment, and investment, referring to both capital formation and natural capital formation. Following the SAM, four types of households are considered for each region according to their income threshold, and who receives income from produc- tion factors and enterprises, as well as who receives income in the form of remittances from abroad and transfers from the 14  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds government. Households also pay taxes to the government and save a proportion of their incomes. Consumer expenditure is a function of prices and incomes according to a Linear Expenditure System (LES) that in its simpler version reduces to fixed expenditure shares and a Cobb-Douglas Utility function (Robinson et al. 1989). Households also spend their incomes to use natural capital, which is added to the expenditure function as an exogenous variable. Intermediate sector demand, including the exchange between the two regions, is given by fixed input-output coefficients. Aggregate spending for government consumption is exogenously determined and defined in terms of fixed shares of aggre- gate government spending for goods and services. Part of government spending is also natural capital, which is added to the government expenditure function and exogenously determined. Sector capital investment is assumed to be allocated in fixed proportions among various sectors and is exogenously determined. The rest of the world includes foreign and out-of-state tourists and is set exogenously. For the balance of trade, we adopt the hypothesis that this is set exogenously and the real exchange rate adjusts to achieve equilibrium. CET, Armington, and export elasticity parameters were taken from literature such as Hinojosa-Ojeda and Robinson (1991), Hanson et al. (1989), and Reinert and Shiells (1991). Factors are assumed to be mobile across activities, available in fixed supply, and demanded by producers at market-clearing prices. Factor incomes are distributed on the basis of fixed shares (derived from base-year data) and transferred to the households. For the depletion of natural capital, an exogenous variable is added to the intermediate use of commodities in the supply-demand equation of final goods. POLICY SIMULATIONS Three critical features of the Kenyan economy become evident. First, investment multipliers for construction The CGE model can be used to determine the effects appear to be linear (i.e., they do not vary with the scale of investment in different sectors by examining the mul- of the investment). Second, the construction multipliers tipliers and the overall impact on key economic vari- are lower than the multipliers associated with conser- ables of interest such as value added (GDP). Table 2.1 vancies and wildlife conservation activities. This is due to and Figure 2.5 show how the CGE multipliers vary with the greater complexity and connectivity of tourism value the size of the investment involved. The table shows chains and the complementarity of wildlife tourism with the consequences of different levels of investment in other sectors of the economy. Third, in contrast to the three sectors—­ construction which proxies investment construction sector, conservation investments exhibit in roads, conservancies as an indicator of investment scale effects and increase with the amount invested. in wildlife tourism, and greater wildlife protection (such These effects emerge as a consequence of deeper link- as anti-poaching patrols and habitat restoration and ages to other parts of the economy. regeneration). Table 2.2 shows the impact on value added of an invest- ment in conservancies of the same magnitude in both TABLE 2.1: CGE investment impact multipliers regions. Since the regions are so different, with the Investment ($, millions) South commanding most of the export trade compared 10 50 100 500 1,000 to the North, an investment in the South has a high own-­ multiplier effect, but virtually no spillover effects to Sector Value added multipliers ($, millions) the North. Investment in the North, on the other hand, Construction 1.96 1.97 1.98 2.06 2.17 spills over into the South. Note too the high investment Conservancies 4.28 4.41 4.57 6.55 13.61 multiplier (4.41) in the North, which reflects the fact that Wildlife 4.26 4.39 4.57 6.75 16.54 investments in wildlife tourism in the North entail bet- Source: Elaboration of the Kenya CGE model. ter utilization of the endowments of land and natural Weighing the Impacts: Generating Scenarios and Simulating Trade-Offs   15 0 500 1,000 1,500 2,000 2,500 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 Agriculture Agriculture Livestock Livestock Forestry Forestry Fishing Fishing Wildlife Wildlife Parks Parks Conservancies Conservancies Water biodiversity Water biodiversity Source: SAM base year data and CGE simqqlations. Mining Mining Food, beverage, and tobacco Food, beverage, and tobacco All other manufacturing All other manufacturing Distribution water Distribution water Utilities Utilities Construction Construction North base case Trade South base case Trade Accommodation and restaurant Accommodation and restaurant Lodge Lodge Transport Transport FIGURE 2.4: CGE model simulation for base-year production activities Information and communication Information and communication Financial and insurance activities Financial and insurance activities Real estate Real estate North simulation Business and administrative Business and administrative South simulation Other services Other services 16  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds Public administration and defense Public administration and defense Health and social work Health and social work Education Education Park tourism Park tourism Beach tourism Beach tourism Cultural tourism Cultural tourism Business tourism Business tourism FIGURE 2.5: Multipliers as a function of investment size 20 15 Multiplier 10 5 0 10 50 100 500 1,000 Investment ($, millions) Construction Conservancies Wildlife Source: Elaboration of the Kenya CGE model. resources, which are relatively abundant in this part of out in the South (Table 2.3). Even though the multipliers the country. are smaller, they are significant in both regions. The size of the multiplier (about 1.5 in both cases) is similar to esti- Consider next an equivalent investment in roads in mates obtained in other countries. both regions. Using the econometric estimates from the previous chapter, it is assumed that this leads to These results suggest that investment in conservan- an expansion in agricultural activity and a reduction in cies and wildlife tourism display important scale effects. wildlife. Investments in conservation and construction both appear to have higher potential in the North, where nat- The absolute impact of investment in construction is ural resources are more abundant and land is cheaper, higher when it occurs in the North. In percentage terms, and where induced tourist activity may spill over to the however, unlike the case of conservancies, investment rest of the country through connections to the better in construction has more balanced results when carried developed southern value chain and infrastructure. TABLE 2.2: Impacts of investment in conservancies (50% of current value = $142 million) Investment in the South (Region A) Investment in the North (Region B) Impact on Region A Impact on Region B Impact on Region A Impact on Region B Value added components ($, millions) ($, millions) ($, millions) ($, millions) Labor 100.12 8.73 96.49 102.95 Capital 120.26 13.34 114.53 175.50 Land 132.69 4.99 22.95 176.92 Other (eco) services 353.08 27.06 233.96 455.36 Total value added 706.15 54.13 467.92 910.73 Investment in the South (Region A) Investment in the North (Region B) Impact on Region A Impact on Region B Impact on Region A Impact on Region B Value added components (%) (%) (%) (%) Labor 0.52 0.51 0.50 6.00 Capital 0.38 0.53 0.36 6.92 Land 2.62 0.56 0.45 20.00 Other (eco) services 6.69 0.64 1.25 25.30 Total value added 0.75 0.54 0.44 10.78 Investment multiplier 3.02 0.22 1.75 4.41 Source: Kenya CGE model. Weighing the Impacts: Generating Scenarios and Simulating Trade-Offs   17 TABLE 2.3: Impacts of investment in construction ($142 million) Investment in the South (Region A) Investment in the North (Region B) Impact on Region A Impact on Region B Impact on Region A Impact on Region B Value added components ($, millions) ($, millions) ($, millions) ($, millions) Labor 91.07 4.31 49.67 79.26 Capital 109.48 6.16 61.87 109.91 Land 11.41 2.02 9.82 15.88 Ecoservices 211.96 12.48 121.35 205.05 Total value added 423.93 24.97 242.70 410.11 Investment in the South (Region A) Investment in the North (Region B) Impact on Region A Impact on Region B Impact on Region A Impact on Region B Value added components (%) (%) (%) (%) Labor 0.47 0.25 0.26 4.62 Capital 0.35 0.24 0.20 4.33 Land 0.23 0.23 0.19 1.80 Other (eco)services 0.25 0.22 0.31 1.89 Total value added 0.38 0.24 0.22 3.75 Investment mutiplier 1.51 0.88 0.1 1.53 Source: Elaboration of the Kenya CGE model. Trade-offs between road construction 77 percent), depending on the amount and location of and wildlife toursim agricultural and livestock expansion. Included are the following three scenarios: (1) a 10 percent increase in Having described the sector multipliers, this section turns investment in road construction and a 15 percent reduc- to the central policy question—the trade-offs involved tion in wildlife in the South; (2) high levels of reduction between road construction and investments in wildlife- in wildlife in the South (30 to 77 percent) as a result of based tourism. To gain a clearer understanding of likely higher levels of road construction; and (3) combining effects and the sensitivity of results to key parameters, conservation and infrastructure policies—capturing the the simulations are based on a wide range of alternative elusive win-wins. scenarios. A variety of cases are considered regarding the sensitivity of tourism to wildlife loss. SCENARIO 1: A 10 percent increase in investment in road construction and a 15 percent reduction in The simulations explore the impact of an increase in wildlife in the South investment in road construction on GDP and its com- ponents, assuming different demand elasticities and The combination of a 10 percent increase in road con- different rates of wildlife loss.7 Clearly, the greater the struction and a 15 percent decrease in wildlife in the sensitivity of tourist demand to wildlife loss, the greater South has a positive impact on agriculture and livestock will be the decline in demand resulting from wildlife production but a generally negative impact on service declines. Likewise, a variety of cases are considered for activities in both the North and the South, but especially the loss of wildlife from road construction (15 and 30, to in the South. The fall in production is particularly large in 7  The CGE model is calibrated using elasticities of substitution of CES the tourism sector. production functions ranging from 0.6 (agriculture) to 1 (industry). CET functions for Armington hypothesis are also calibrated with a higher elasticity range (0.5 to 2) and elasticity of foreign tourism demand with respect to wildlife The impact on value added in the South (Table 2.4) mainly ranging from 0.3 to 1.5. The model is run with a Keynesian closure with labor occurs through land, whose demand rises because of the supply perfectly elastic, capital mobile across each region, and wage as the numeraire. expansion of agriculture and trade. In the North, on the 18  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds TABLE 2.4: Impact on regional value added of an increase in infrastructure (+ 10%) and reduction in wildlife in the South (percent from baseline) South (Region A) Wildlife (15%) Wildlife (15%) Wildlife (15%) Tourism demand Tourism demand Tourism demand elasticity = 1 elasticity = 0.6 elasticity = 0.3 Labor 3.71 3.81 3.90 Capital 3.41 3.49 3.57 Land 12.78 12.99 13.20 Ecoservices –1.10 –0.80 –0.49 Total value added 4.21 4.31 4.41 North (Region B) Labor 4.60 4.71 4.82 Capital 5.80 5.92 6.03 Land 4.74 4.89 5.03 Ecoservices –3.36 –3.19 –3.03 Total value added 4.39 4.51 4.63 Source: Elaboration of the Kenya CGE model. TABLE 2.5: Impact on regional income distribution of an increase in infrastructure and reduction in wildlife in the South (percent from baseline) South (Region A) Wildlife (15%) Wildlife (15%) Wildlife (15%) Tourism demand Tourism demand Tourism demand elasticity = 1 elasticity = 0.6 elasticity = 0.3 Enterprises 3.41 3.49 3.57 Rural poor 5.37 5.49 5.61 Rural non-poor 5.30 5.42 5.54 Urban poor 3.46 3.55 3.64 Urban non-poor 3.44 3.53 3.62 North (Region B) Enterprises 5.80 5.92 6.03 Rural poor 3.95 4.06 4.17 Rural non-poor 4.00 4.11 4.22 Urban poor 4.21 4.31 4.42 Urban non-poor 3.84 3.94 4.03 Source: Elaboration of the Kenya CGE model. other hand, income increases across all factors of produc- In conclusion, when the decline in wildlife and tour- tion. In both regions, multipliers are high, indicating both ism demand elasticity is moderate, there is an overall direct and indirect effects of the same orders of magni- increase in value added (GDP), with the decrease in the tude and large spillovers from backward linkages. The tourism value chain being compensated by the increase results suggest an overall improvement across most sec- in value added in other parts of the economy. It is also tors of the economy, despite the loss of tourism income. useful to note that the poor in both rural and urban areas Weighing the Impacts: Generating Scenarios and Simulating Trade-Offs   19 benefit equitably in this scenario. In this case, it pays to South benefits and the North contracts. In the first two deplete the natural assets that attract tourists, since the scenarios where wildlife is assumed to decrease 30 per- gains from other sources of income outweigh the losses. cent, total value added in the South increases. In the North, however, land incomes fall in response to the SCENARIO 2: High levels of reduction in wildlife higher supply of more accessible and fertile lands in the in the South (30 to 77 percent) as a result of higher South, and value added is reduced. This is a scenario in levels of road construction which more amenable conditions in the South “crowds out” economic activity from the North. A second set of simulations investigates the same investment in road construction but with higher impacts The third and fourth columns explore scenarios with on wildlife in the South, involving reductions of wild- higher rates of wildlife reduction (assumed to be 70 per- life biomass 30 to 77 percent in both regions. Elastici- cent according to recent trends in areas close to roads). ties of tourism demand with respect to wildlife are also In this scenario both regional economies suffer, with neg- assumed to be higher, ranging from 1 to 1.5. ative value added changes being especially large in the North. The losses accruing from the decline in tourism In both regions there is a general boost to the economy in revenue and the associated value chain outweigh any the agricultural and construction sectors. But in all cases gains that a road might bring. What is especially striking considered, this is insufficient to compensate for the fall is the magnitude of the loss in the North relative to the in production that is catalyzed by the near collapse of the South, reflecting the different comparative advantages wildlife tourism industry and its value chain. Moreover, of the two regions. since these effects are the result of spillovers from the South, the multiplier effects are similar in both regions, The value added effects bring to light a central disconti- with only a slight tendency for the North to compensate nuity in the response of the economy, which is illustrated with its larger and cheaper supply of land and labor. in Figure 2.6. In this diagram, the size of the balls rep- resent the assumed elasticity of tourism demand with In terms of value added however, differences emerge respect to wildlife, while the horizontal and vertical axes across regions and scenarios (Table 2.6) where the measure the changes in value added and the reduction TABLE 2.6: Impact on regional value added of an increase in investment in road construction (+10%) and greater reduction in wildlife in the South (percent from baseline) Wildlife (30%) Wildlife (30%) Wildlife (77%) Wildlife (77%) Tourism demand Tourism demand Tourism demand Tourism demand elasticity = 1 elasticity = 1.5 elasticity = 1 elasticity = 1.5 South (Region A) Labor 4.13 3.76 –0.05 –0.52 Capital 4.50 4.17 1.21 0.80 Land 10.35 9.63 –5.69 –6.54 Ecoservices –9.18 –10.07 –34.63 –35.66 Total value added 4.62 4.23 –0.54 –1.02 North (Region B) Labor 1.52 1.13 –8.83 –9.30 Capital 2.55 2.14 –8.82 –9.29 Land –5.28 –5.75 –30.15 –30.62 Ecoservices –15.73 –16.22 –44.78 –45.25 Total value added –1.08 –1.50 –16.26 –16.73 Source: Elaboration of the Kenya CGE model. 20  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds FIGURE 2.6: Value added variation in regard to wildlife reduction and tourism elasticity 0% –10% –20% Reduction in wildlife –30% –40% –50% –60% –70% –7 – –80% 8 –8 –90% –3.00% –2.00% –1.00% 0.00% 1.00% 2.00% 3.00% 4.00% 5.00% 6.00% Value added variation Source: Kenya CGE model. in wildlife, respectively. The diagram shows that out- FIGURE 2.7: Relationships between wildlife reduction comes cluster around two key points: (1) a moderate and value added, and value added growth level of reduction of wildlife with low tourism elasticity 60,000 and an associated increase in value added, and (2) a 59,500 Value added ($, millions) high level of reduction of wildlife, with an associated fall 59,000 58,500 in value added. 58,000 57,500 To summarize, the simulations suggest that when the 57,000 decline in wildlife is modest (less than around 30 per- 56,500 cent), then the benefits of construction investments out- 56,000 weigh the losses brought by a decline in tourism and its 55,500 value chain. The North is more vulnerable to the adverse 55,000 0% 20% 40% 60% 80% 100% impacts due to limited alternative forms of economic Reduction of wildlife activity. But when the decline is large (≈ 70%), there is a fundamental shift in the balance of costs and benefits. In 2.50% this case the loss of income associated with the panoply of wildlife tourism–related value chains, outweighs the 2.00% benefits from improved access from road investments. 1.50% Value added growth 1.00% While it may be objected that these are hypothetical 0.50% simulations, the estimates are based on observed mag- nitudes, suggesting these results are a cause for policy 0.00% consideration. Box 2.2 provides a more detailed expla- 0% 20% 40% 60% 80% 100% –0.50% nation of these results in the context of a production pos- –1.00% sibility frontier. –1.50% Reduction of wildlife Figure 2.7 shows the relationship between wildlife reduc- Source: Elaboration of the Kenya CGE model. tion and GDP emerging from these model solutions, with a general equilibrium frontier exhibiting an inverted “U” shape pattern. The figure represents a production pos- gain with value added increasing. Beyond this thresh- sibility frontier. It shows that when construction induces old, further declines in wildlife induce net losses of value declines in wildlife that are relatively modest and less added. The current magnitude of wildlife loss across than around 30 percent, then there is a net economic much of the country suggests that Kenya is on the Weighing the Impacts: Generating Scenarios and Simulating Trade-Offs   21 BOX 2.2:  Trade-offs between Economic Growth and Environmental Impact The CGE model summarizes an economy-wide equilibrium outcome that is termed a “general equilibrium” (GE). The results of the model can be used to define the outcome and trade-offs between economic growth and ecological effects. The curve in Figure B.2.1 summarizes the outcomes of the simulations conducted in this exercise. It shows that at low levels of environ- mental impact, growth rises with environmental deterioration but it then reaches a turning point and begins to decline after around a 30 percent loss of wildlife. FIGURE B2.1: Impact on environmental deterioration against GDP growth 6% 5% 4% GDP growth 3% 2% 1% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% Environmental deterioration This curve in fact represents an efficiency frontier, in the sense that it bounds a feasible set of growth rates and degrees of environmental deterioration (ED). Points below the curve are both feasible and inefficient. In the first part, for example, a combination of 3 percent GDP growth and 20 percent ED could be improved upon by increasing growth at the same level of ED or by reducing ED and maintaining the same level of growth. In the second part of the curve, this would also be possible by exploiting the two branches of the curve. For example, the combination of 1 percent growth and 60 percent ED could be improved upon by increasing growth up to the declining branch with the same amount of ED, or by decreasing ED with the same amount of growth by moving to the increasing branch of the curve. The latter case, however, would signal a much larger inefficiency than the former one. More generally, the non-monotonic relationship between economic and ecological outcomes, popularized as the Kutznets curve, suggests that growth-depressing feedback may indefinitely prolong the negative relationship between develop- ment, inequality, and a deteriorating environment. For example, the limits theory (Arrow et al. 2013) defines the economy-­ environment relationship in terms of environmental damage hitting a threshold beyond which production is so badly affected that the economy shrinks. The so-called new toxics view claims that emissions of existing pollutants are decreasing with economic growth, but the new pollutants substituting for them increase with growth. In fact, consistent with the new toxics hypothesis, the U.S. EPA claims that it receives premanufacturing notices to approve over 1,000 new chemicals each year. declining portion of this frontier. This is a region where appear to be the population group most disadvantaged good conservation becomes good economics. by the negative effects on the tourism industry, espe- cially in the North. This is unsurprising as the evidence The impact on income distribution reflects, to an extent, on conservancies presented in Chapter 5 suggests that the changes in value added, and is also highly asymmet- wildlife tourism provides employment to sections of the ric across regions and income groups (Table 2.7). In spite labor market with low levels of human capital and few of the surge in agriculture in all scenarios, the rural poor fungible skills. 22  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds TABLE 2.7: Impact on income distribution of an increase in infrastructure and a greater reduction in wildlife in the South (percent from baseline) Wildlife (30%) Wildlife (30%) Wildlife (77%) Wildlife (77%) Tourism demand Tourism demand Tourism demand Tourism demand elasticity = 1 elasticity = 1.5 elasticity = 1 elasticity = 1.5 South (Region A) Enterprises 4.50 4.17 1.21 0.80 Rural poor 5.24 4.80 –1.79 –2.33 Rural non-poor 5.25 4.81 –1.60 –2.14 Urban poor 4.07 3.71 0.07 –0.38 Urban non-poor 3.97 3.61 –0.22 –0.66 Government 4.51 4.15 –0.92 –1.37 North (Region B) Enterprises 2.55 2.14 –8.82 –9.29 Rural poor –0.15 –0.53 –12.68 –13.11 Rural non-poor 0.00 –0.38 –12.36 –12.80 Urban poor 1.77 1.38 –7.45 –7.88 Urban non-poor 1.78 1.43 –6.58 –6.99 Source: Elaboration of the Kenya CGE model. TABLE 2.8: Doubling the investment in conservancies: impact on value added Value added South (Region A) North (Region B) % change ($, millions) Base case Simulation Base case Simulation Region A Region B Labor 19,324.90 19,500.00 1,764.40 1,764.90 0.9 0.0 Capital 31,278.20 31,497.90 2,607.40 2,634.20 0.7 1.0 Land 5,163.20 5,292.00 895.40 911.30 2.5 1.8 Ecoservices 1,214.00 1,289.30 700.30 701.90 6.2 0.2 Source: Elaboration of the Kenya CGE model. SCENARIO 3: Combining conservation and infra- distribution is regionally unbalanced, however, with the structure policies—capturing the elusive win-wins investment boosting overall economic activities in the North, but with most benefits spilling over to the South. A third set of simulations assesses the possible con- Natural capital activities (maintenance and conservation) sequences of win-win policies, i.e., policies aimed at increase in both regions. increasing (doubling) investment by targeting both envi- ronmental preservation and efficiencies. For this purpose, When investments in conservancies are also com- three components of possible investment policies were plemented with wildlife preservation,8 the results in analyzed: (i) expanding conservancies, (ii) preserving Tables 2.10 and 2.11 suggest a synergic effect, with a high wildlife, and (iii) increasing productivity through “smart” beneficial impact (investment multiplier = 2.42), which infrastructure of the kind described in the next chapter. would favor a pattern of growth more balanced across regions and income groups. The simulations indicate As Tables 2.8 and 2.9 show, doubling the investment 8  Wildlife preservation includes all investment aimed at identifying, protecting, in conservancies has an overall positive effect (invest- and expanding key areas to help wildlife thrive, and in many cases, recover ment multiplier = 1.9 in terms of total value added). Its from endangered and threatened status. Weighing the Impacts: Generating Scenarios and Simulating Trade-Offs   23 TABLE 2.9: Doubling the investment in conservancies: impact on income distribution Income South (Region A) North (Region B) % Change ($, millions)   Base case Simulation Base case Simulation Region A Region B Enterprises 31,278.20 31,497.90 2,607.40 2,634.20 0.7 1.0 Rural poor 7,996.50 8,102.30 1,683.90 1,698.30 1.3 0.9 Rural non-poor 13,069.50 13,238.20 2,607.00 2,629.40 1.3 0.9 Urban poor 1,437.90 1,450.40 214.80 216.10 0.9 0.6 Urban non-poor 36,365.70 36,693.50 6,480.90 6,523.90 0.9 0.7 Investment in conservancies 285.4 570.8 0.1 0.2 Source: Elaboration of the Kenya CGE model. TABLE 2.10: Doubling investment in conservancies and wildlife conservation: impact on value added Value Added South (Region A) North (Region B) % Change ($, millions) Base case Simulation Base case Simulation Region A Region B Labor 19,324.90 20,884.90 1,764.40 2,503.30 8.1 41.9 Capital 31,278.20 33,931.20 2,607.40 4,005.90 8.5 53.6 Land 5,163.20 7,722.90 895.40 2,218.30 49.6 147.7 Ecoservices 1,214.00 2,374.70 700.30 1,964.00 95.6 180.5 Source: Elaboration of the Kenya CGE model. TABLE 2.11: Doubling investment in conservancies and wildlife conservation: impact on income distribution South (Region A) North (Region B) % Change Income ($, millions) Base case Simulation Base case Simulation Region A Region B Enterprises 31,278.20 33,931.30 2,607.40 4,005.90 8.5 53.6 Rural poor 7,996.50 9,615.10 1,683.90 2,772.90 20.2 64.7 Rural non-poor 13,069.50 15,642.00 2,607.00 4,261.90 19.7 63.5 Urban poor 1,437.90 1,580.70 214.80 303.30 9.9 41.2 Urban non-poor 36,365.70 40,401.30 6,480.90 8,903.90 11.1 37.4 Investment in conservancies 285.4 570.8 0.1 0.2 Investment in wildlife 1,598.8 3,197.5 529.7 1059.3 Source: Elaboration of the Kenya CGE model. that agriculture and livestock would contract (moder- such example of a new technology. It is a protected area ately) in the South and expand in the North, where the management tool designed to measure, evaluate, and economy would grow in terms of both value added and improve the overall effectiveness of law enforcement personal incomes. patrols.9 In this simulation, the model predicts synergis- tic effects with more than proportional increases of the Tables 2.12 and 2.13 show the results of the simulations of multipliers. The impact on incomes is large and more a hypothetical scenario that involves combining “smart” balanced across regions and income groups, with the technologies with traditional conservation techniques North and the poor reaping the largest benefits. In sum, through productivity increases and resource allocation. The Spatial Monitoring and Reporting Tool (SMART), 9 https://loisaba.com/smart-using-cutting-edge-technology-monitor-loisabas- already in use in some conservancies in Kenya, is one wildlife-populations/ 24  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds TABLE 2.12: Doubling investment and capital productivity in conservancies and wildlife conservation: impact on value added Value Added South (Region A) North (Region B) % Change ($, millions) Base case Simulation Base case Simulation Region A Region B Labor 19,324.90 22,029.30 1,764.40 3,102.80 14.0 75.9 Capital 31,278.20 36,626.40 2,607.40 5,476.40 17.1 110.0 Land 5,163.20 9,358.40 895.40 3,284.50 81.3 266.8 Ecoservices 1,214.00 3,142.90 700.30 2,993.00 158.9 327.4 Source: Elaboration of the Kenya CGE model. TABLE 2.13: Doubling investment and capital productivity in conservancies and wildlife conservation: impact on income distribution Income South (Region A) North (Region B) % Change ($, millions) Base case Simulation Base case Simulation Region A Region B Enterprises 31,278.2 36,626.5 2,607.4 5,476.4 17.1 110.0 Rural poor 7,996.5 10,769.4 1,683.9 3,702.5 34.7 119.9 Rural non-poor 13,069.5 17,492.1 2,607 5,676.2 33.8 117.7 Urban poor 1,437.9 1,701.1 214.8 383.8 18.3 78.7 Urban non- poor 36,365.7 43,777 6,480.9 11,055.3 20.4 70.6 Investment in conservancies 285.4 570.8 0.1 0.2 Investment in wildlife 1,598.8 3,197.5 529.7 1,059.3 Source: Elaboration of the Kenya CGE model. “smart” investments in conservation could be a “win- and ecological balance has also undermined the pro- win” policy with huge gains for both regions, a healthy ductivity of natural capital. The overall effects of these balanced expansion of the economy, and larger benefits trends has resulted in a negative link between population for the rural poor. growth and agricultural expansion on the one hand, and the productivity of renewable natural capital on the other. Rural poverty and tourism On the positive side, a significant reduction of rural pov- Rural poverty and the conservation of natural capital are erty has occurred because the pattern of development linked to each other in several ways. First, a majority of the in Kenya has been sufficiently diversified to offer both rural poor directly depend for their livelihoods on agricul- alternative and complementary economic opportunities ture, pastoralism, and other natural resource–dependent to the rural populations. In the past 20 years, Kenya has livelihoods. Second, this dependence, while supporting developed a diversified industrial and service economy, their subsistence status, is also risky, as it exposes them with a vibrant tourism industry, which is itself diversified to the vagaries of weather and the oscillations of market across the whole range of the country’s considerable prices. Third, because the population continues to grow supply of alternative products, from beaches to land- at very high rates, the pressure on land increases and scapes rich in wildlife. Nature-based tourism thrives on productivity (per person) tends to fall, making the plight of a value chain directly dependent on local agriculture, pastoralists and small farmers facing a shrinking resource agroindustry, and specialized services. base ever more dramatic. Because landholdings are sub- divided across an increasing population, the expansion The development of tourism in Kenya is thus a part of the of agriculture at the expense of traditional pastoralism transformation from quasi-subsistence into commercial Weighing the Impacts: Generating Scenarios and Simulating Trade-Offs   25 agriculture and brings with it greater integration into the rest rural poor in the South to the urban poor in the North, of the economy. Increasing reliance on the market has sev- but their basic values are not very different across the eral dimensions, including the share of consumption that is various income groups. purchased in the market, expenditure for food as a share of total expenditure, old and new sources of off-farm income, If all tourism-related activities (not just expenditures debt, and the need for storage facilities. Tourism-related ser- on maintaining parks) are given a boost by increasing vices and employment provide a series of backward link- investment in parks and conservancies, as shown in ages that increase the flexibility of the farm household in Figure 2.8, income elasticities (percentage increases in ordinary times, reduce its direct and exclusive dependence incomes in response to 100 percent increase in spend- on agricultural markets, and make the poor more resilient ing) rise significantly (ranging from 25 percent to 16 per- to adverse shocks. The backward linkages of tourism to the cent) and the difference in response between rural and rural economy may thus improve income prospects and sta- urban and poor and non-poor groups is heightened. For bility for all the rural population, including the rural poor. completeness Figure 2.8 and Figure 2.9 also show how these elasticities vary between urban and rural areas. The CGE captures the interdependence between the rural economy and nature-based tourism, both through the estimates of transactions across the value chains, Concluding comments and through the regional and economy-wide multipli- The CGE model developed for this study presents a ers arising from backward and forward linkages. Gen- picture of the Kenyan economy, with stark differences eral equilibrium price effects are also estimated by the of factor supply and employment between the more model, which registers a rise in value added through developed South and the less developed North. The both higher factor employment and higher prices of land. two regions are interdependent to an extent, especially because most of the industrial and service value added For example, Figure 2.8 shows how in the CGE base is produced in the South. These linkages result in invest- solution, household income elasticities, with respect to ments in the North generating larger spillovers in the park tourism expenditure (i.e., the percentage increases South in absolute terms, following a pattern common to in incomes following a 100% increase in park tourism many unequal regional economies. At the same time, for expenditure), range from 3.4 percent to about 2.3  per- activities that depend on open spaces and nature, dam- cent across income groups in the two regions. The elas- age to wildlife and tourism value chains in the South tend ticities decline smoothly from their highest value for the to negatively affect both regions. However, absolute FIGURE 2.8: Income elasticities Income elasticities with respect to park tourism expenditure (ESAM base values from multipliers) 3.5% 3.0% 2.5% Elasticities 2.0% 1.5% 1.0% 0.5% 0.0% h h h h h th th th ut ut ut ut t or or or or So So So So N N N N r or or or r or or or o o po po po po po po po po n- n- n- n- al n al n ba ba r r no no no no Ru Ru Ur Ur al n al n ba ba r r Ru Ru Ur Ur Household income 26  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds FIGURE 2.9: Income elasticities Income elasticities with respect to conservation tourism activity level (ESAM base values from multipliers) 30.0% 25.0% Elasticities 20.0% 15.0% 10.0% 5.0% 0.0% h h h h th th th th ut ut ut ut or or or or So So So So N N N N r or or or r or or or oo oo o po po o po po lp -p lp -p n- n- n n on on ra ra ba ba no no Ru Ru ln ln Ur Ur n n ra ra ba ba Ru Ru Ur Ur Income groups effects are larger in the South, while relative damages REFERENCES are proportionally higher for the nascent tourism activi- ties in the North. Arrow, K. J., P. Dasgupta, L. H. Goulder, K. Mumford, and K.  Oleson. (2013).  “Sustainability and the Measurement of Wealth: Further Reflections.” Environment and Devel- The present surge of infrastructure investment in Kenya opment Economics 18 (4). is thus likely to bring some benefits to the already devel- Hanson, Kenneth, Sherman Robinson, and Stephen Tokarick. oped regions, though this will come at a cost of increasing (1989). Working paper no. 510 United States Adjustment congestion and aggravating inequalities and environmen- in the 1990s: A CGE Analysis of Alternative Trade Strat- tal damage. Where the damage is large, it could outweigh egies. California Agriculture Experiment Station, Univer- the benefits. The reason is that the decline in wildlife sity of California, Berkeley. results in a drastic fall of nature-based tourism in both Hinojosa-Ojeda, R., and S. Robinson. (1991). “Alternative sce- regions, as well as a decline in many service sectors due narios of US-Mexico integration: A computable general to the linkages. Perhaps of greater concern is that these equilibrium approach.’ Working paper series, California impacts are disproportionately felt by the rural poor and Agricultural Experiment Station, Department of Agricul- in the North. Prospects of development in this region tural and Resource Economics. thus appear to be vulnerable to investment choices in the Perali, F., and Scandizzo, P. L., eds. (2018). The New Gen- eration of Computable General Equilibrium Models, South because of the concentration of economic activities Springer. in this more developed region and the widespread nega- Reinert, K. A., and C.R. Shiells. (1991).  Trade Substitution tive effects on the environment and tourism in the North. Elasticities for Analysis of a North American Free Trade Area. US International Trade Commission. In sum, if wildlife reduction is large (which it is now), Robinson et al. (1989). Multisectoral models. In Handbook and/­or demand elasticities of tourism are high (which is of development economics, vol. II, ed. H. Chenery and also probably true), higher investment in infrastructure T. N. Srinivasan. Amsterdam: Elsevier Science Publishers. may lead Kenya into a development trap, where major Sanghi, Apurva, Richard Damania, Farah Manji, and Maria negative effects on wildlife, the environment, and tour- Paulina Mogollon. (2017). Standing out from the herd: ism ultimately hamper both its resources and its eco- An economic assessment of tourism in Kenya (English). nomic growth. The empirical findings suggest that, at the Washington, D.C., World Bank Group. present, Kenya is moving closer to this trap, which it will Scandizzo, Pasquale L., and Ferrarese, Cataldo. (2015). likely only escape by appropriately combining invest- “Social Accounting Matrix, a New Estimation Methodol- ogy,” Journal of Policy Modeling 37 (1). ment in both infrastructure and conservation policies. Weighing the Impacts: Generating Scenarios and Simulating Trade-Offs   27 CHAPTER 3 WILDLIFE-FRIENDLY ROADS: FABLE OR FACT? The development of new roads in Kenya will be crucial to GIS data have significantly expanded the scope of such spurring growth and human development, and promot- analyses, notably in data-poor contexts (Iimi et al. 2016). ing shared prosperity. At the same time, as established The approach based on GIS data was refined and scaled in previous chapters, the expansion of Kenya’s road to 166 countries by Mikou et al. (2019), who also devel- network ranks high in the list of factors contributing to oped a tool to help predict which all-season roads should wildlife loss. As the CGE analysis has indicated, where be built by upgrading existing tracks in order to maximize wildlife losses are substantial, the economic benefits the RAI. Indeed, an algorithm using information pertain- that a road brings may not outweigh the benefits for- ing to where the population lives, where all-season roads gone, especially in the more remote parts of the country exist, and where other roads/tracks are located can lead where economic opportunities are limited (and multipli- to prioritizing road improvements that connect the highest ers are small). This problem would be overcome if it were number of people to the network at the lowest cost. possible to construct a road with limited impact on wild- life in ways that minimize losses and maximize benefits. This chapter applies the method pioneered by Mikou This chapter presents a new tool that can help identify et al. (2019) to the Kenyan context and goes a step further which roads should be developed based not only on to take into account the externalities generated by the their economic potential, but also factoring in the pos- road network. The method relies on data of human pop- sible negative impacts on wildlife. The results highlight ulations, existing roads, and a set of possible new roads. the existence of important margins to develop an eco- WorldPop data from 2015 provides gridded estimates of nomically inclusive road network that at the same time population distributions at a 1-kilometer resolution, and acknowledges externalities and is respectful to wildlife. similar to the methodology outlined in Chapter 1, data for Kenya’s existing major roads are derived from Michelin maps (2017). DeLorme data for Kenya is used to identify New approaches to enhancing paths and tracks that are potential candidates for new road access roads (Figure 3.1). The DeLorme dataset is considered to be comprehensive and up-to-date regarding transpor- Roads are key for economic development, and as pre- tation infrastructure, including roads, paths, and tracks, viously highlighted, a staggering 70 percent of Kenya’s but it is limited in terms of information on the quality of rural population still lives more than 2 kilometers from surfacing. Michelin data are then used to more precisely an all-season road. The SDGs promote the construc- classify which segments correspond to existing all- tion of all-season roads, defined as roads motorable all weather roads and which ones correspond to paths or year round by the prevailing means of rural transport. In tracks. The latter are then used as candidate segments the relatively dry context of Kenya, in addition to tarmac for possible extensions of the road network. roads, paved and improved roads are also considered as all-season roads. Indicator 9.1.1 of the SDGs encourages policy makers to increase the share of the rural population Conventional approaches living within a 2-kilometer distance of an all-season road, calculated as the Rural Access Index (RAI). Earlier studies of increasing road access first measured RAI using household survey data (Roberts Conceptually, a road network is a mathematical graph. et al. 2006), but advances in technology and the use of This graph can be extended by converting a track 28  FIGURE 3.1: Existing all-weather roads and tracks in Kenya Source: Michelin maps (roads); DeLorme data (tracks). connected to the current network into a road. An algo- wildlife are not internalized or considered in the con- rithm determines all possible new graphs that would struction process. be formed by the connection of one new track to the existing graph. For each graph, the new RAI is calcu- According to available data, the share of Kenya’s rural lated. By doing so, the algorithm determines which population living within 2 kilometers of an all-season track leads to the highest increase in the RAI, and road is currently around 28 to 30 percent. The algorithm based on the length of each segment, it determines developed here suggests that this RAI could be increased the cost of converting this segment into a road. Here, to more than 50 percent simply by converting existing the construction cost of the road is a linear function of tracks to roads. Figure 3.2 displays the marginal and total the length of the new road (see Mikou et al. 2019 on cost of increasing the RAI, expressed as a percentage costs). The track that brings the maximum increase of of GDP. The cost of increasing the RAI is fairly constant the RAI at the lowest cost is chosen and added to the from the current level up to about 45 percent of the popu- road network. A more complete mathematical graph is lation. The figure indicates that for about 2.5 percent of consequently formed, and the procedure is repeated current GDP, an additional 15 percent of the rural popula- until no gain in the RAI is possible. This method from tion could be connected to the road network. This addi- Mikou et al. is used to determine a set of priority roads. tional 15 percent roughly represents 6 million new people It adopts what could be termed a “business as usual” who primarily live in Kenya’s densely populated western scenario in which the negative effects of roads on counties and around Nairobi (Figure 3.3). Wildlife-Friendly Roads: Fable or Fact?   29 FIGURE 3.2: The costs of increasing Kenya’s RAI under the “business as usual” scenario Marginal cost of increasing RAI 2.0 50 45 1.5 Cost (% of GDP) RAI (% ) 40 1.0 35 0.5 30 0.0 30 35 40 45 50 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 RAI (%) Cumulative cost (% of GDP) Source: Michelin and DeLorme data; Method developed by the World Bank. FIGURE 3.3: Building new roads to increase Kenya’s RAI, starting with the densely populated western counties Source: Michelin and DeLorme data; Method developed by the World Bank. Progressively, more remote areas start to be con- GDP—which is the same cost as connecting the first nected to the network. However, the cost of con- 15 percent of the population to the network. This is necting each additional household sharply increases; consistent with global trends observed by Mikou et al. for instance, increasing the RAI from 45 percent to (2019) across Sub-Saharan Africa. However, thanks to 46 percent would cost an additional 0.5 percent of the higher GDP of Kenya compared to most other Afri- GDP. Even more so, bringing the RAI to 52 percent can countries, the relative cost of increasing its RAI, (from 51 percent) would cost a further 2.5 percent of expressed in GDP, is lower. 30  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds Counting the costs of business Figure 3.4 shows how much wildlife would be lost as the RAI increases. The results suggest that the costs for wild- as usual life associated with extending the road network slowly What would the environmental cost of the “business as begin increasing and are followed by losses of wildlife usual” scenario be? Of primary interest, the western part sharply increasing as more people are connected to the of Kenya,10 where many roads would be upgraded, falls network. Even among the first road segments built in outside the rangelands and is home to limited wildlife. the rangelands, important wildlife areas are threatened. This suggests that a large part of the rural population Observe that the impact on wildlife is constant for the first could be connected to the road network at a low envi- 1.5 million people connected to the network, and that it ronmental cost in terms of biodiversity loss. jumps very dramatically thereafter. The CGE analysis in Chapter 2 warns that losses of this scale bring adverse To quantify the impact on wildlife when this conventional GDP consequences, especially in areas with limited “business as usual” approach is used, biomass data of potential for growth and labor-intensive employment. ungulate wildlife, derived from DRSRS, are overlapped with roads. Using the estimates presented in Chapter 1, it is assumed that the conversion of a track into a road A greener scenario would lead to a decline of wildlife in a 20-kilometer buf- However, even if the costs outweigh the benefits of such fer around the newly built road. At each step of the algo- policies outlined above, it is unlikely that this would pre- rithm, wildlife loss in each extension is calculated. vent the construction of roads in the rangelands. This FIGURE 3.4: Wildlife loss is constant for the first 1.5 million people connected to the road network, with losses sharply increasing thereafter 250,000 200,000 Impacted wildlife (kg) 150,000 100,000 50,000 0 0 500,000 1,000,000 1,500,000 2,000,000 2,500,000 3,000,000 Population with new access to all-weather roads Source: Authors. 10  This would include the following counties: Migori, Homa Bay, Kisii, Nyamira, Bomet, Kericho, Kisumu, Nandi, Vihiga, Siaya, Busia, Bungoma, Trans-Nzoia, Marakwet, Uasin Gishu, Nakuru, Nyandarua, Nyeri, Muranga, and Nairobi. Wildlife-Friendly Roads: Fable or Fact?   31 FIGURE 3.5: The costs of increasing Kenya’s RAI under the two scenarios 38 36 34 RAI (%) 32 30 0 0 2 4 6 8 10 12 14 Cumulative cost (% of GDP) Normal RAI RAI with wildlife constraint section demonstrates that a more careful extension of The results are promising. The first striking finding is that the road network allows for as many people to be con- both models (the original one as well as the model with nected to the network as in the “business as usual” sce- the added parameter on wildlife impact) attain the same nario, at a similar cost, but with moderate consequences increase in the RAI at comparable cumulative costs (Fig- for wildlife. ure 3.5). When focusing only on the rangeland counties for which there is biodiversity data, the current RAI of To run the greener scenario, the original algorithm from about 28 percent could be increased with both models Mikou et al. (2019) was modified, allowing for the con- to approximately 38 percent. This holds for the model sequences of road construction on wildlife to be con- that does not include a wildlife constraint (green line) as sidered. Thus far, the objective function of the algorithm well as the modified model that factors in a wildlife con- was to maximize the number of people connected to straint (orange). the network at the lowest cost. In this section, an extra parameter is added: simultaneously minimizing the Of crucial importance, the model that includes the wild- impact on wildlife. As is standard in statistical analysis, life constraint allows for a significant reduction in the loss human population data and wildlife biomass data were of wildlife from increased road access. Figure 3.6 com- normalized and scaled over the same support to ensure pares the environmental effectiveness of both models, that neither one was overweighed in the algorithm.11 In highlighting that the modified model (orange line) offers doing so, the objective of this approach was to find areas solutions to connecting people to the road network while where roads could be built to maximize access and mini- having limited detrimental effects on wildlife. mize impact on wildlife. Under the original model, wildlife is lost after approxi- mately 500,000 people are connected to the road network (green line), while wildlife loss in the modified model only happens after 2 million people gain access 11  Mathematically, the objective function of the algorithm is: maximizing people/ (km*wildlife impacted). to these improved roads. Further, while wildlife loss in 32  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds FIGURE 3.6: Factoring in wildlife constraints significantly reduces the impact of new roads on wildlife 200,000 175,000 150,000 Impacted wildlife (kg) 125,000 100,000 75,000 50,000 25,000 0 0 500,000 1,000,000 1,500,000 2,000,000 2,500,000 3,000,000 Population with new access to all-weather roads Current situation More wildlife friendly roads the first model skyrockets after 1.5 million people are source of information for this model (Figure 3.7).12 In addi- connected, under the modified model, this happens tion, though the model built in this instance was trained after 2.5 million people gain access to the road network. to prioritize road improvement in order to connect the Hence, most people could be connected to the network highest number of people to the network, a similar model while avoiding negative impacts on wildlife. could be adjusted to connect the area with the highest agricultural potential to the network, or areas with the highest poverty rates to the network. This would con- Fine-tuning the model stitute a fine-tuning of the model but would not change the central message: huge opportunities exist to extend The results presented above come with a few caveats. Kenya’s road network and to protect wildlife at the same More than definitive results, the value of this exercise time. lies in its original approach—developing a tool that could be used to inform decision making and to understand In sum, smarter, greener approaches to infrastructure the trade-offs between wildlife protection and economic are also economically more beneficial. Achieving this opportunities. The model developed could also be fur- outcome is not impossible, and it requires policy mak- ther refined to provide more fine-tuned policy messages. ers to properly identify areas where roads should not be The protection of wildlife corridors has become a critical constructed. aspect for wildlife protection in Kenya, as most are under intense threat of conversion for other land use. Similar to the way wildlife density data were introduced into the 12  Among other refinements, we should note the possibility of varying the model, data on wildlife routes could also be included. functional form of the objective function of the algorithm—the size of the buffers built around each road for which we assume an impact on wildlife (here Precise information on these routes is being gathered by 20 kilometers). It could be 10 kilometers in a “more aggressive scenario” or leading experts in Kenya and could become a valuable 30 kilometers in a “more conservative scenario.” Wildlife-Friendly Roads: Fable or Fact?   33 FIGURE 3.7: Mapping elephant and wildebeest routes in Kenya Source: Ojwang et al. (2017). 34  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds REFERENCES Ojwang, Gordon O., et al. (2017). Wildlife Migratory Corridors and Dispersal Areas: Kenya Rangelands and Coastal Ter- Iimi, A., F. A. K. Ahmed, E. C. Anderson, A. S. Diehl, L. Maiyo, restrial Ecosystems. T. Peralta Quiros, and K. S. Rao. (2016). New Rural Access Roberts, P., S. KC, and C. Rastogi. (2006). Rural Access Index: Index  : Main Determinants and Correlation to Poverty. A Key Development Indicator. Washington, D.C. The World Washington, D.C. The World Bank. Bank. Mikou, M., J. Rozenberg, E. Koks, C. Fox, and T. Peralta Quiros. (2019). Assessing Rural Accessibility and Rural Roads Investment Needs Using Open Source Data. Washington, D.C. The World Bank. Wildlife-Friendly Roads: Fable or Fact?   35 CHAPTER 4 THE WAY FORWARD AND NEXT STEPS A 70 percent decline in wildlife, within thirty years, is a Realizing this economic potential will call for a significant sobering statistic. As Kenya’s population grows, its infra- shift in two key policy areas. First, it will require changes structure needs expand, and climate change makes in the way in which intrusive infrastructure is planned rainfall more erratic, and the pressures on wildlife and and located to avoid the fragmentation and conversion natural habitats will intensify in regions that are already of natural habitats with economic potential. Second, under environmental stress and will spread to other parts there is a need to create the enabling conditions to real- of the country. The journey along the current policy path ize the economic potential through investments in con- has failed to halt the degradation and fragmentation of servancies at scale. Neither approach will be sufficient natural habitats, and it is unlikely to do so in the future on its own and both will need to work in tandem: the first when pressures expand and competition for land, water, to prevent the loss of economic opportunities by land and other natural resources intensifies. This suggests conversion, and the second to harness economic poten- an urgent need for a careful reassessment of pressures, tial through investments. The remainder of this chapter policies, and future prospects. discusses critical elements of this approach. Wildlife in Kenya, especially in the North of the country, rep- resents a lucrative economic asset whose contribution has Smart infrastructure been underestimated and potential unrealized. The CGE Where ecotourism potential exists, it is important that assessment indicates that every dollar invested in conser- infrastructure investments are done with consideration of vation and wildlife tourism could generate benefits that ecotourism’s impacts on these assets. The fact that much range from $3 to $20. For comparison, it is instructive to remains to be built creates an opportunity to build “right.” note that in the United States and Brazil, $1 invested in pro- Getting infrastructure “right” is critical because infra- tected areas generates approximately $6–$8 as a return structure choices have long-lived and difficult-to-reverse (do Val Simardi Beraldo Souza, 2017). Table 4.1 illustrates impacts on land, wildlife, water, and future patterns of that in Kenya the economic benefits from investments in development. Infrastructure decisions influence the type wildlife tourism rise with the amount that is invested. Such and location of development and, as such, create sub- increasing returns likely reflect the ecological importance stantial inertia in economic systems, with irreversible con- of connected natural habitats that are more productive in sequences that need to be weighed against alternatives. terms of the ecosystem services that they provide and are also more resilient to droughts and other weather extremes Recognition of these complex issues suggests the need (Haddad et al. (2015). In the remote and arid North of the for a different approach to infrastructure needs with a country there are few other investments that could yield a focus on “building right” rather than simply “building more.” comparable economic return. Building right typically brings benefits that accrue over the longer term. The fact that infrastructure needs are so large TABLE 4.1: GDP multipliers for investments implies that there are wide opportunities to build right— (in million USD) in conservancies garnering benefits while minimizing or avoiding possible negative impacts on the country’s comparative advantage. 10 50 100 500 1,000 Investment in North 3.13 3.16 3.19 3.52 4.02 The right infrastructure also offers substantial co-­benefits conservancies South 5.43 5.63 5.89 9.07 20.2 that could enhance the productivity and earning capacity 36  of the country’s natural capital. The trade-offs and syner- same tools can be used to predict how to meet other gies from infrastructure and roads are considerable and development objectives more effectively. Through warrant closer examination in decision making. This is careful and strategic planning, spending on infra- especially important for remote parts of Kenya with a structure can be rendered more effective and more limited natural comparative advantage for arable agri- conducive to growth and poverty reduction, and less culture. Where appropriately managed, there are con- impactful on wildlife and the economic opportunities siderable synergies between wildlife tourism and cattle that they bring (Box 4.1). The additional complexity ranching, both of which offer climate resilient livelihood and cost of planning, such as in infrastructure, would opportunities in areas with limited economic potential. be justified by the vastly greater benefits that would As human population densities increase throughout accrue to the country. Africa, there will be a growing premium on places that offer such experiences. Destroying this economic poten- tial could be a short-sighted strategy. Realizing economic opportunities through conservancies Development of large strategic infrastructure to pro- mote connectivity can be consistent with efforts to Conservancies could play a crucial role in halting the col- conserve natural assets, which also contribute to eco- lapse of wildlife in Kenya by extending the areas under nomic growth. As illustrated in Chapter 3, tools are protection around parks, reconnecting habitats, and available that allow planners to predict the impacts of limiting overcrowding in parks. And more than that, con- their decision on wildlife—a key economic asset. The servancies offer levers to boost and diversify economic BOX 4.1:  Smart Infrastructure and Spatial Planning The lack of spatial planning when combined with inadequate investment in infrastructure can create dynamics that are unsustainable and non-inclusive. There are significant deficiencies with the piecemeal and project-by-project assessment of each investment alternative in isolation. One obvious consequence is that options which generate higher benefits may be overlooked since the focus is on a single project. Another and seldom recognized problem is that of “dynamic inconsistency”: where the first project unleashes conse- quences for other projects. For instance, suppose that the first project diminishes environmental quality in a protected area. This makes it more likely that another intrusive structure will “pass” a cost-benefit test. The first project therefore unleashed a dynamic that leads to complete transformation of the landscape, which was not considered at the outset. This is termed dynamic inconsistency and leads to poor decision making and economically unwarranted destruction of natural assets. Against this background of escalating and suboptimal land conversion, two new concepts of spatial planning are advanc- ing, both require prioritizing ecosystem services (forests, rural areas, watersheds, urbanized vast areas, etc.). One approach uses physical measures in GIS models to avoid damage and build synergies with ecosystems, as illustrated in Chapter 3. The other takes a more economic approach by adopting a set of values or shadow prices that make the land use scale hier- archical and compatible with the functionality of potential ecological networks. This requires prioritizing ecosystem services (forests, rural areas, watersheds, etc.) by adopting a set of values or shadow prices that make the land use scale hierarchical and compatible with the functionality of potential ecological networks. Combined with higher capacity for project manage- ment, implementing the new concept of infrastructure is a promising strategy to invest wisely and more effectively. In sum the idea is to make aspirations for “smart” infrastructure into a reality by using tools to combine functional efficiency, technology, and ecosystem conservation. The Way Forward and Next Steps  37 activities in some of the most remote parts of the coun- The contribution of conservancies to the tourism industry try. In places where ranching and agriculture are under remains modest—it accounts for a meager 1.3 percent of stress due to shifting weather patterns, land degradation, earnings in the industry, suggesting considerable poten- or overstocking, conservancies offer more sustainable tial and scope for expansion in a specialized market that livelihood options that will inevitably increase in value as caters to the high-value and low-volume tourists. A sur- wildlife numbers and wilderness viewing opportunities vey of 13 regional associations and 160 conservancies shrink across the globe. In sum a strategic expansion of registered under KWCA suggests that there are around conservancies offers an opportunity to complement the 2,510 beds available in lodges within conservancies, government’s current focus and most (97 percent) are found in the southern con- servancies. The average conservancy in the sample has More generally, conservancies represent projects that 28 beds, but with considerable variation ranging from offer a platform to integrate ecological and economic 6 in Machakos, with its conservancies being in the early functions, which contrasts with the segregated conven- stages of development, to over 1,000 in Narok, which tional approaches of conservation and development. abuts the overcrowded Maasai Mara. By allowing an array of organizational forms based on the coexistence of activities involving agriculture, live- Tourism is the most significant source of income for stock, conservation, and different forms of culture and conservancies, contributing an average of 83 percent nature-based activities, conservancies widen the menu of commercial income with buoyant growth in recent of choices and offer a promising strategy to end the years.13 Cattle ranching has, over the past years, gained chaotic process of landscape fragmentation and wildlife prominence and offers a way to diversify income extirpation. sources. A key challenge is to keep livestock herds in balance with wildlife numbers in cultural contexts where There are currently around 160 conservancies in Kenya, livestock is more than an economic asset. Iconic animal spread across 28 counties, under the umbrella of Kenya conservation programs (of species such as rhino, ele- Wildlife Conservancies Association (KWCA). These cover phant, Grevy’s zebra, chimpanzee) and other payment around 11 percent of the country’s territory, with 3.7 mil- for environmental service programs are also a significant lion hectares in the North and 2.1 million hectares in the contributor to incomes, with conservancies earning an South (Figure 4.1). By comparison, the terrestrial national average of Kenya shilling (Ksh) 12.8 million in 2016 and parks and reserves cover 4.7 million hectares. Ksh 11 million in 2017 from conservation fees. Conservancies significantly increase the share of wild- For communities who live within or near conservancies, life living in legally protected areas. Around 22 percent there are significant benefits. The survey indicates that of the total ungulate wildlife biomass is found in conser- the 160 conservancies hired around 2,600 people, and vancies. This represents a significant complement to the provide bursaries and educational support especially to 38 percent of ungulate wildlife biomass found within women, and are a significant source of income for food Kenya’s national parks. Perhaps of greater importance, and other provisions required by tourists. Income from 18 out of 20 zones with the highest wildlife density are conservancies is the only drought-proof source of rev- found in conservancies rather than in parks. For exam- enue that is available to many of the poor and vulnerable ple, Olare Orok, located next to the Maasai Mara, is the communities. conservancy with the highest density of wildlife biomass. Key species such as the Grevy’s zebra are mostly found Despite these benefits, investments in conservancies in conservancies, while lion populations in the conser- carry high risks and as such require patient capital. This vancies of the Maasai Mara are among the highest on is because investors must gamble not only on the pros- the continent (Elliot and Gopalaswamy 2017; Ogotu et al. pects of attracting tourists to a new location, but must also 2017). The data suggest that there is a lag in the recovery engage in a host of investments to build community sup- of ungulate biomass in conservancies with the greatest port and fill crucial infrastructure gaps. This may suggest increase occurring in conservancies that were created in the 1980s (Figure 4.2). 13  NRT, 2018, State of Conservancies Report, 2017. 38  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds FIGURE 4.1: Map of parks and conservancies in Kenya (2018) Source: Authors. the need for innovative investment mechanisms such as is not sufficient to assure success. The establishment and green bonds and risk guarantees to shift the risk-reward promotion of conservancies offers the most scalable ave- balance, especially in areas that confer high ecological nue in ensuring wildlife habitats are secured and migra- benefits. Recognizing that the conservancies confer pub- tion corridors are established. Wildlife hot spot areas, lic benefits, there is a case to be made for enabling policy such as the Mara, Amboseli, and Laikipia regions, indicate support—for example through investments in marketing that high wildlife densities can lead to significant wildlife- strategies aimed at both local and international travelers. based tourism operations outside of national parks. The presence of wildlife in conservancies has been the To further promote the development of tourism outside single most important determinant of success, though this of national parks and reserves, the national and county The Way Forward and Next Steps  39 FIGURE 4.2: Wildlife generally increased in the older conservancies and decreased in areas where conservancies were established after 1995 600 550 Percentage change in wildlife, 1980–2010 500 450 400 Dyanmic 350 Gain Loss 300 250 Size impact 200 200 400 150 600 100 50 0 –50 –100 1970 1980 1990 2000 2010 Year of establishment of conservancy Source: Authors using DRSRS data and conservancies data developed in this report. governments need to recognize the role conservancies Haddad, N. M., Brudvig, L. A., Clobert, J., Davies, K. F., Gonzalez, play as custodians of wildlife and in developing syner- A., Holt, R. D., . . . and Cook, W. M. (2015). Habitat fragmenta- gistic livelihood enhancement programs. Integration of tion and its lasting impact on Earth’s ecosystems.  Science conservancy management plans in the county develop- Advances 1(2), e1500052. ment plans acts as a first step to foster this recognition. Ogutu, Joseph O., et al. (2017). “Wildlife population dynamics in human-dominated landscapes under community-based conservation: the example of Nakuru Wildlife Conservancy, Kenya.” PloS one 12.1: e0169730. REFERENCES Olare Orok Conservancy.” Master’s thesis, University of Oslo. do Val Simardi Beraldo Souza, T. T. (2017). Economic impacts of tourism in protected areas of Brazil. Journal of Sustainable Tourism, 1–15. Elliot, Nicholas B., and Arjun M. Gopalaswamy. (2017). “Toward accurate and precise estimates of lion density.” Conserva- tion Biology 31.4: 934–943. 40  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds APPENDIX A CONSERVANCIES—AN OVERVIEW The history of conservancy development in Kenya was largest number of conservancies, each hosting 25 con- founded upon conservation practices introduced by the servancies, with Narok (16) and Nakuru (14) following suit. British colonialists in the 1800s and 1900s. These altered The northern counties of Samburu, Isiolo, Marsabit, Tur- the traditional land tenure system and enabled commer- kana, Garissa, and Mandera host a much smaller share cial harvesting of wildlife, leading to significant declines of Kenya’s conservancies (23), while 19 counties located in wildlife numbers. The 1933 “London Convention” rep- in the Central and Western regions of Kenya do not host resented a turning point that marked the beginning of any conservancies at present the end to commercial wildlife harvesting, and it vested authority to a central body for wildlife management. In Of the 160 conservancies, 107 are currently operational, 1946, the National Park Ordinance resulted in the estab- 44 are emerging, and 9 are proposed. As seen in Fig- lishment of Nairobi, Tsavo, Mount Kenya, and Aberdares ure A.1, the three types of conservancies found in Kenya National Parks. Game hunting and an increase in human- include (i) community conservancies—those set up by a wildlife conflict in the 1950s and 1960s led to the cen- community on community land for the purpose of liveli- tralization of wildlife management. Non-state protected hood development and wildlife conservation; (ii) private areas—as they were called before the term conservan- conservancies—those set up on private land by a private cies was coined—emerged at this time, with the creation individual or corporate body for the purpose of wildlife of the Solio, Ol Jogi, Sangare, Sergoit, and Taita Hills pro- conservation; and (iii) group conservancies—those which tected areas for rhinos and other wildlife species. include the creation of a single conservancy by private landowners who pool land for the purpose of wildlife Momentum for conservancies gained traction in the conservation. 2000s with the formation of regional conservation groups such as The Northern Rangelands Trust (NRT) Community conservancies first appeared in Kenya in and the South Rift Association of Landowners (SORALO). the mid-1990s with support from nonprofits, neighbor- The establishment of a national association in 2012—the ing private conservancies, and conservation-oriented Kenyan Wildlife Conservancies Association (KWCA)— corporations as a way of incentivizing landowners and helped to further promote the approach.14 communities to be custodians of wildlife. The success of establishing Kimana in 1992, Namunyak and Koiyaki- There are currently more than 160 conservancies in Lemek Wildlife Trust in 1995, and Il Ngwesi in 1996, all Kenya, spread across 28 counties, under the umbrella of which offered direct economic benefits from wildlife- of KWCA. The overwhelming majority of these conser- related activities to landowners, catalyzed the growth vancies (137) are located in the country’s South, with of the community conservation model (Figure A.1). The Kajiado and Taita Taveta counties being home to the establishment of group conservancies in the south- ern counties in the 2000s was catalyzed by the need 14  This was driven by the draft Wildlife Conservation and Management to create wildlife dispersal areas and ensure con- Bill of 2011 and the Conservancy Regulations of 2012, which both explicitly recommended devolution of rights to landholders and the institutionalization of nectivity of subdivided lands outside the Maasai Mara the wildlife industry in Kenya (Kenya Wildlife Conservancies Association, “Our National Reserve and the Amboseli National Park. This Story,” https://kwcakenya.com/about-us/our-story/). A study tour to the Namibian Association of Community Based Natural Resource Management Support also created an opportunity to sell an exclusive wild- Organization (NACSO), consultative meetings with over 600 stakeholders, life experience to visitors, promoting high-end, low- followed by a national consultative forum, enabled the endorsement and registration of KWCA in December 2012 and April 2013, respectively. impact safari-based tourism, an alternative to the mass   41 FIGURE A.1: The rapid growth of conservancies in Kenya 160 Conservancy type Community Group 140 Private 120 Number of conservancies 100 80 60 40 20 0 1965 1970 1977 1984 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 Source: KWCA Conservancy Database, 2018. tourism experience in the neighboring national parks. TABLE A.1: Typology of Kenyan conservancies Now classified as a conservancy, Sergoit Farm was the Number and % Area Area first privately owned area set aside for the conservation Conservancy type of conservancies (ha) (%) of rhinos outside of national parks and reserves in 1953. Community conservancy 82 (51%) 6,100,000 76 This was followed by Ol Jogi in 1965 and Wangalla Ranch in 1968. Following the hunting ban in the 1980s, other Private conservancy 58 (36%) 1,200,000 15 private entities turned to a combination of ranching and Group conservancy 20 (13%) 723,000 9 conservation, driving the growth of private conservan- Note: Analysis is based on a sample of 130 conservancies assessed for this study. cies in the Taita Taveta, Laikipia, and Rift Lakes regions up until the mid-2000s. The environmental promise The majority of Kenya’s conservancies (51 percent) are on community land, while 36 percent have been established of conservancies on private land, and 13 percent exist on group lands Conservancies span more than 11 percent of Kenya’s ter- (Table A.1). Because of the ability of wildlife and livestock ritory, over 5.8 million hectares, with the northern conser- to coexist, coupled with the expanse of conservancies vancies covering 3.7 million hectares and the southern and connectivity between them, communally owned pas- conservancies covering 2.1 million hectares. By com- toral lands host vast amounts of wildlife in Kenya. This parison, Kenya’s terrestrial national parks and reserves has, by default, led to community conservancies offering cover 4.7 million hectares, spanning 16 counties. As the significant conservancy potential and demonstrating the country develops, conservancies can play a significant largest growth in the conservancy movement. role in securing a place for wildlife in Kenya’s future. 42  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds Conservancies significantly increase the share of wildlife potential. Conservancies now offer key possibilities to living in legally protected areas, A spatial assessment of extend and differentiate Kenya’s tourism product. For the biodiversity indicates that 22 percent of the total ungulate first time, this report has collected data on the economic wildlife biomass is found in conservancies, according to contribution of conservancies through tourism (Box A1). DRSRS data. This represents a significant complement to the 38 percent of ungulate wildlife biomass found within The 13 regional associations and 160 conservancies Kenya’s national parks. Perhaps of greater importance, registered under KWCA were surveyed in 2018 to col- 18 out of 20 zones with the highest wildlife density are lect information on Kenya’s tourism infrastructure and found in conservancies rather than in parks. Olare Orok, sources of income of conservancies in 2016 and 2017. located next to the Maasai Mara, is the conservancy with Twenty-five tour operators were also approached to the highest density of wildlife biomass. Key species such gather data on income paid to conservancies, bed- as the Grevy’s zebra are mostly found in conservancies, nights, benefit sharing mechanisms, and philanthropic while lion populations in the conservancies of the Maa- activities supported within the conservancies. sai Mara are among the highest on the continent. These figures highlight the crucial role conservancies can play in protecting wildlife and helping landscapes thrive. BOX A.1:  Some Key Figures on the More significantly, a growing body of evidence suggests Economics of Conservancies in Kenya that conservancies have been highly successful at pro- • More than 930,000 members in conservancies tecting biodiversity. For instance, in Nakuru Wildlife Con- servancy, Ogotu et al. (2017) found that populations of • 131 tourism facilities (~2,500 beds) monitored wildlife in the conservancy had stabilized for • 175,000 bed-nights in 2017, a 30% increase com- some species and increased for most, in stark contrast pared to 2016; occupancy of 20%. to the declines observed elsewhere, including in the • 2,620 locals directly employed (20% women) national parks. • Tourism operators paid more than Ksh 1.2 billion in bed-nights to conservancies in 2017 The economic significance of conservancies Of the 160 conservancies documented in this study, Kenya’s tourism sector generated Ksh 99.7 billion in 2016, 69 host a total of 131 tourism facilities within their borders a figure that increased by 20.3 percent to Ksh 119.9 bil- (Table A.2). A total of 2,510 beds exist in lodges within lion in 2017 (KNBS 2018). According to the Kenya Wildlife the conservancies mapped, with 97 percent found in the Service (KWS) Strategic Plan of 2012–2017, safari tour- southern conservancies. Of the total beds, 41 percent ism accounts for 75 percent of national tourism earnings are located in Narok County, 13 percent each in Kajiado (Ksh 74.8 billion in 2016 and Ksh 90 billion in 2017). But and Laikipia counties, 11 percent in Taita Taveta, and the share of tourism income earned by conservancies 9 percent in Nakuru. The Mara conservancies (located in amounted to a modest 1.3 percent, suggesting consider- Narok County) currently host the largest number of facili- able potential and scope for expansion in a specialized ties outside of national parks and reserves (37 percent). market that most likely caters to the high-value and low- It should be noted though that the scope for expansion volume tourists. of tourism activity is constrained by a limit on “bed- nights” (conservancies such as Olare Orok only allow a Safari tourism—first established as hunting safaris and single bed per 300 acres) (Bedelian 2014). These limits progressing to ecotourism—has been one of the top rev- are meant to assure an exclusive game viewing experi- enue earners for Kenya, with national parks historically ence and build a differentiated market and product to playing a crucial role. It has offered income-generating the high-volume tourism in the parks. Table A.2 provides prospects to pastoral households in the arid and semi- an overview of the scale of tourism operations in the arid regions of Kenya, which are areas of low agricultural conservancies surveyed. Conservancies—An Overview  43 TABLE A.2: An overview of tourism facilities in Kenya’s TOURISM: THE PRIMARY SOURCE OF INCOME conservancies FOR CONSERVANCIES No. of Average no. Tourism is the most significant source of income for No. of tourism of beds per conservancies, contributing an average of 83 percent County conservancies facilities conservancy of commercial income (NRT 2018). There are signs that Baringo 2 2 15 income from tourism is growing rapidly in relative and Elgeyo Marakwet 1 1 Under absolute terms. From 2016 to 2017, the 69 conservan- construction cies with tourism facilities experienced an 18 percent Kajiado 11 13 28 increase in their total income from tourism, earning a total Laikipia 11 23 31 of Ksh 1.15 billion (Figure A.2). This amounted to an aver- Lamu 1 1 21 age of Ksh 26.2 million per conservancy (a minimum of Machakos 2 2 3 Ksh 20,000 and a maximum of Ksh 253 million). Growth in income was highest in the northern conservancies, Meru 2 7 57 who saw a 33 percent increase in tourism income, com- Nakuru 9 15 26 pared to a 23 percent increase among conservancies in Narok 14 49 76 the south. Nyeri 1 1 24 Samburu 4 6 16 Members of conservancies (i.e., the local households) Taita Taveta 7 7 39 share the benefits from tourism either directly as reve- Tana River 1 1 14 nues from running tourism facilities or through a matrix of profit-sharing structures, conservation fees, bed-night Trans Nzoia 1 1 32 fees, or lease-holding arrangements (Box A.2). Vihiga 1 1 20 West Pokot 1 1 12 Total 69 131 28 BOX A.2:  Types of Benefit-Sharing Arrangements In general, conservancies that neighbor highly fre- quented parks and reserves have higher average bed Bed-night fee: A proportional fee paid per occupied bed densities, as they take advantage of other attractions to the conservancy. and better access. The average conservancy in the sam- Lease-holding fee: A set monthly or annual fee paid out ple has 28 beds, with numbers ranging from 6 in Macha- as rent for land or building infrastructure for an agreed- kos, with its conservancies being in the early stages of upon period. development, to over 1,000 in Narok, which abuts the overcrowded Maasai Mara. The Amboseli and Laiki- Conservation fee: An additional fee paid per visitor or pia regions are also wildlife hot spots, with a proximity occupied bed as a payment for conservation services. to Mt.  Kilimanjaro and Mt. Kenya adding an additional attraction for visitors. The bulk of income to conservancies (> 50 percent) Conservancies target the high-value international tour- is generated by fees earned from tourism-related ist, though there are a growing number of local visitors benefit-sharing agreements, followed by livestock sales. ­ with much regional variation. In Nakuru County, about Noncommercial activities, animal conservation (12 per- 60 percent of visitors are local, and in Narok and Taita cent), and payments for ecosystem services (8 percent), Taveta the figure stands at 30 percent, which is close to together with livelihood activities, account for the rest of the national average. On the other hand, in Laikipia, Sam- the income (Figure A.3). buru, Meru, and Kajiado, the percentage of local tourists is much lower (at around 15 percent)—these being des- The expansion of cattle ranching and improved beef tinations that are targeted to the international traveler. production has, over the past years, gained prominence 44  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds FIGURE A.2: Tourism income earned by conservancies (Ksh, 2017) North Samburu 51,099,560 Isiolo 172,500 Total 51,272,060 South Narok 489,152,444 Laikipia 342,591,509 Meru 151,938,077 Nakuru 62,212,694 Kajiado 50,741,184 Taita Taveta 2,508,824 Tana River 350,000 Baringo 681,000 Nyeri 109,280 Vihiga 60,000 Total 1,100,345,012 Grand total 1,151,617,072 North South Grand total Source: Conservancies surveyed in this study. FIGURE A.3: Proportion of conservancy income sources in 2017 (Ksh, millions) 18 17 124 8 136 214 Bednight and conservation fee Tourism leasehold 1172 Livestock sales 323 Payment for ecosystems Iconic conservation Farming Grazing fee Bead sales 492 Hay sales Source: Conservancy survey. within community and group conservancies, as the Iconic animal conservation programs (of species such need for wildlife-compatible opportunities arises. as rhino, elephant, Grevy’s zebra, chimpanzee) are also However, this remains a challenge as grazing regimes a significant contributor to incomes, with conservan- need to be established, and an equilibrium between cies earning an average of Ksh 12.8 million in 2016 and livestock and wildlife-carrying capacities needs to be Ksh 11 million in 2017 from conservation fees paid by determined and managed, keeping in mind cultural visitors to animal sanctuaries. Iconic animal conservation contexts of livestock being a measure of wealth within programs, which were the initial drivers of conservancy these communities. development in the 1960s, continue to attract tourists. Conservancies—An Overview  45 Payments for ecosystem services, particularly from car- this is not sufficient to assure success. Critically, there bon sequestration, have increasingly become an impor- is a need for strong governance structures with trans- tant revenue source for conservancies. In 2016, southern parent and equitable benefit-sharing structures. Invest- conservancies earned Ksh 30.4 million from carbon off- ments in conservancies carry high risks and as such sets (an average of Ksh 4.4 million per conservancy), and require patient capital. This is because investors must this figure increased by 605 percent in 2017 to reach gamble not only on the prospects of attracting tourists to Ksh 214.4 million (an average of Ksh 21.4 million per con- a new location, but also engage in a host of public good servancy). This was mainly due to carbon-offset revenues investments to build community support and fill crucial from the Chyulu Hills REDD+ project, a multi-partner ini- infrastructure gaps. This may suggest the need for inno- tiative aimed at reducing emissions from deforestation vative investment mechanisms, such as green bonds and degradation. As the international policy framework and risk guarantees, to shift the risk-reward balance, around land-based climate change strategies continues especially in areas that confer high ecological benefits, to mature, landscape-level conservation will offer oppor- such as wildlife corridors. tunities to reap benefits from payments for ecosystem services. Though tourism is the primary income-generating source for most conservancies, accounting for almost OTHER BENEFITS TO COMMUNITIES 83  percent of income (NRT 2018), conservancies and their regional associations are exploring ways to inno- Tourism facilities within the conservancies hired vate and create income from other sources. The Chyulu 2,111  employees (12 percent women) in 2016 and Hills REDD+ project has demonstrated returns at scale 2,619 employees (17 percent women) in 2017. Most conser- from conservation through payments from ecosystem vancies are located in pastoral areas where gender ineq- services. While cattle ranching also offers opportuni- uity exists in terms of access to education and economic ties, it is more complex in the context of degraded land, opportunities, with traditional livelihood practices limiting increasing population numbers, and the need to balance women’s opportunities outside the homestead. However, livestock numbers with wildlife populations due to lim- as gender empowerment through bursary and education ited carrying capacity. support continues to be promoted through conservancy management structures, this trend may change. The southern tourist circuit in Kenya hosts a well-­ maintained infrastructure and offers opportunities for tour- The facilities also provide alternative sources of income ists to travel by road within a radius of one to five hours to households through direct purchases of goods and from Nairobi. It also hosts high wildlife densities and ben- services, which amounted to around Ksh 36.5 million efits from strong marketing. Such potential also exists in in 2017, cultural activities such as visits to homesteads destinations such as Laikipia and in the North more gener- (Ksh 11.9 million in 2017), the purchase of livestock and ally, which also host some of the highest wildlife numbers food (Ksh 36 million in 2017), and the purchase of bead- in the country. This region, however, requires significant work (Ksh 4 million). Tourism facilities have also invested investments in marketing strategies aimed at both local in roads, education, health, and water-related infrastruc- and international travelers. Critically, as other chapters in ture in some of the most remote regions of the country. In this report have highlighted, there is also a need for infra- 2017, 11 conservancies had invested about Ksh 28.6 mil- structure approaches that carry a lower negative footprint lion in such activities, suggesting that the unaccounted in order to catalyze and enable the economic opportuni- impact of tourism in the form of social initiatives may be ties that Kenya’s natural assets bring. more significant than direct payments to conservancies in the form of tourism operations. Going further PUTTING THE NUMBERS IN PERSPECTIVE Establishing stable or increasing wildlife population num- The presence of wildlife in conservancies has been the bers is critical toward enhancing tourism income, with its single most important determinant of success, though potential for addressing high poverty in rural areas. The 46  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds establishment and promotion of conservancies offers tourism experiences, need to be incorporated into the most scalable avenue in ensuring wildlife habitats the country’s parks and reserves plans to achieve are secure and rehabilitated, and migration corridors are the national goal of the country becoming a premium established. Wildlife hot spot areas, such as the Mara, destination of high-end safari tourism. The triple bot- Amboseli, and Laikipia regions, indicate that high wildlife tom line of conservation, livelihoods, and economic densities can lead to significant wildlife-based tourism sustainability provided by conservancies should be operations outside of national parks. marketed as a unique wildlife experience within this portfolio. There is also a need to promote conservan- In addition to this, the assessment of policies and pro- cies through Kenya Tourism Board (KTB) and Ministry grams across all sectors that impact wildlife numbers of Tourism programs. should be established to ensure for wildlife-friendly national development plans. REFERENCES To further promote the development of tourism outside Bedelian, Claire. (2014). Conservation, Tourism and Pastoral of national parks and reserves, the national and county Livelihoods: Wildlife Conservancies in the Maasai Mara, governments need to recognize the role conservancies Kenya. play as custodians of wildlife and in developing syner- Jenya National Bureau of Statistics (KNBS). (2018). Kenya Eco- gistic livelihood enhancement programs. Integration of nomic Survey 2018. conservancy management plans in the county develop- Northern Rangelands Trust (NRT). (2018). “State of Conservan- ment plans acts as a first step to foster this recognition. cies Report, 2017.” Furthermore, financial support to strengthen proposed Ogutu, Joseph O., et al. (2017). “Wildlife population dynamics and growing conservancies on their path to sustainabil- in human-dominated landscapes under community-based ity will catalyze growth of the movement. conservation: the example of Nakuru Wildlife Conservancy, Kenya.” PloS one 12.1: e0169730. In line with Vision 2030, conservancies, which have already paved the way for exclusive wildlife-based Conservancies—An Overview  47 APPENDIX B ROAD EXTENSION AND WILDLIFE LOSS BETWEEN 1980 AND 2010: A DIFFERENCE-IN-DIFFERENCES APPROACH In this appendix, we detail the methodologies and the To analyze how wildlife population has changed over results of the econometric model that estimate the time and spatially the data were aggregated into ­census impact roads had on wildlife in Kenya between the periods covering surveys conducted between ­1977–1989 1980s and the 2000s. Results echo the findings of a sig- (1980s), 1990–1999 (1990s) and 2000 and 2016 (2000s). nificant amount of literature including previous work by For each grid, population estimates were calculated the World Bank, that shows a negative effect of roads on based on biomass (calculated in terms of Tropical Live- natural habitats—notably forests. stock Units where 250 kg is equivalent to 1 TLU) for the 18 common wildlife species17 and were averaged for each of the counting periods. The average over the time Data period minimizes the influence of stochastic variation in the count totals and the distribution of animals. WILDLIFE Wildlife data come from the Department of Resource Sur- WILDLIFE DYNAMICS veys and Remote Sensing (DRSRS) of Kenya based on aer- ial surveys in the rangelands of Kenya since 1977. DRSRS In the 1980s, wildlife was present in 53 percent of grid cells. conducted a total of 359 surveys from 1977 to 2016 cover- In the 2000s, this number is of only 31 percent (Figure B.1). ing 19 rangeland counties. Each county is partitioned into The densities of wildlife in the 1980s were highest in the 5 km × 5 km UTM grids. Each 5-km transect segment is southern rangelands, and the northern rangelands also had treated as an observation unit. Systematic transect lines substantial wildlife distributed across the northern range- are flown through the center of each grid on a north-south land counties. The highest wildlife densities in the 1980s or east-west axis at a nominal height of 91–122 m (300 to were observed in the counties of Narok, Kajiado, Taita, 400 feet) aboveground. Widths of counting strips ranged Lamu, and Laikipia. The 2000s distribution map indicates between 224–490 m during 1977–2016. Two rear-seat that the wild herds have shrunk in numbers and distribu- observers count all wild and domestic animals the size of tion, and have vanished rapidly in many counties including Thomson’s gazelle (15 kg) and larger within each strip and West Pokot, Turkana, Baringo, Kilifi, Lamu, Machakos, and record all counts on tape recorders. Animals in large herds Tana River (Said et al., 2016).18 of more than 10 are photographed and later counted under a binocular microscope (in earlier years) or on a large digi- ROADS tal screen (currently) in digital photos. Refer for details to Norton-Griffiths (1978)15 and for survey parameter (survey Kenya’s road network has grown considerably over dates, aircraft settings, sampling fraction, and personnel the last decades. We use Michelin maps of East Africa involved) to Ogutu et al. (2016). Population estimates (PE) and their standard errors (SE) for each species are calcu- 17  Eighteen species are used in the analysis of the report: buffalo (Syncerus caffer), Burchell’s zebra (Equus burchelli), Coke hartebeest (Alcelaphus lated from the sample fraction by treating each transect as buselaphus), eland (Taurotragus oryx), elephant (Loxodonta africana), gerenuk a sample unit using Jolly’s Method 2 (Jolly, 1969).16 For com- (Litocranius walleri), giraffe (Giraffa cemelopardalis), Grant’s gazelle (Gazella granti); Grevy’s zebra (Equus grevyi), impala (Aepyceros melampus), lesser kudu putation limits, data were resampled at a 10 km afterwards. (Tragelaphus imbermbis), oryx (Oryx gazelle beisa), ostrich (Struthio camelus), Thomson’s gazelle (Gazella thomsoni), topi (Damaliscus lunatus korrigum), warthog (Pharcoerus africanus), waterbuck (Kobus ellipsiprymnus), and 15  Norton-Griffiths, M. (1978). Counting animals. Nairobi: Africa Wildlife wildebeest (Connochaetes taurinus). Leadership Foundation. 18  Said, M. Y., Ogutu, J. O., Kifugo, S. C., Makui, O., Reid, R. S, and de Leeuw, J. 16  Jolly, G. M. (1969). Sampling methods for aerial censuses of wildlife (2016). Effects of extreme land fragmentation on wildlife and livestock population populations. East African Agricultural and Forestry Journal, 34, 46–49. abundance and distribution. Journal for Nature Conservation, 34: 151–164. 48  FIGURE B.1:  Kenya’s wildlife populations have shrunk dramatically since the 1980s, becoming fragmented, and almost vanishing in some counties, such as in West Pokot, Baringo, Turkana, Machakos, Kwale, and Mandera to highlight these changes and study the impact of from a road. Simultaneity bias may be a significant threat road expansions on wildlife. For this study, all avail- when studying the impact of roads on wildlife since wild- able Michelin maps for Kenya were digitized and trans- life distribution and road placement are jointly deter- formed into GIS files. In 1978, the maps recorded about mined. Difference-in-differences models are an effective 7,000 kilometers of paved and improved roads, and the method to overcome this challenge. entire north of the country only featured improved gravel roads at the time. In the subsequent 40 years, Kenya’s Cells that were originally (1980s) far from a road road network has increased by 50 percent to cover (50–100  km) are kept in the analysis. Among these ­ around 11,000 kilometers of improved and paved roads cells, the model looks at how the loss of wildlife differed as of 2017. The network of roads has become denser in between cells that became closer to a road (treatment the South but has also been extended in the North to groups, 5 km, 10 km, 51 km, 20 km, and 50 km to test for connect the major urban center in the region, an exam- the robustness of the estimates) and cells that remained ple being the recent paving of roads leading to Marsabit far from a road (control group, >50 km). Roads here include and Turkana counties. both paved and improved roads. Formally, the model is: Wildlifei,t = β Cell Close from Roadi,t + γ Posti,t + ω Cell The model Close from Road ∗ Posti,t + µt × Province + PAi,t + ∈i,t A “difference-in-differences” specification is used to Where Wildlifei,t is the total biomass of wildlife in cell i determine the impact of roads on wildlife loss. It follows during decade t (t = 1980, 1990, 2000), Cell Close from best practices, followed by recent studies such as Asher, Road measure whether the cell has become 5, 10, 15, Garg, and Novosad (The Economic Journal, forthcom- 20, or 50 km closer to a road during the period. Post ing). The model exploits the expansion of the road net- is a dummy variable for periods post 1980s (i.e., once work in Kenya in the 1980s–1990s. most cells became close to a road). The interactive term Cell Close from Road ∗ Posti,t captures the difference-in- The Euclidean distance between each grid cell and the difference impact of roads on wildlife. µt × Province is a nearest paved or improved road was calculated for each province specific time fixed effect. PAi,t is a time varying decade from the 1980s to the 2000s. These distances variable that equals one if the cell belong to a Protected were then categorized into different bins depending on Area during a given decade. Cells at the borders of Kenya whether a cell was less than 5, 10, 15, 20, or 50 kilometers have a smaller area than cells which do not touch the Road Extension and Wildlife Loss between 1980 and 2010: A Difference-in-Differences Approach   49 border. Therefore, observations are weighted regarding DIFFERENCE-IN-DIFFERENCE ESTIMATES the area of each cell. Finally, standard errors are clus- Table B.1 presents results of the main model approach. tered at the cell level to account for heteroskedasticity. Results from the statistical model suggest that cells located close to a road are associated with a significant The main results are presented in Table B.1. In addition decrease in wildlife, following construction of the road. to showing the robustness of the results to different dis- The closer a cell is to a road, the larger the impact. Results tance thresholds, we also show their robustness in the in Table B.1 reveal that a cell that was once 50 kilome- standard parsimonious difference-in-difference model: ters away from a road, and which subsequently had a Wildlifei,t = β Cell Close from Roadi,t + γ Posti,t + road built less than 5 kilometers away from it, lost an ω Cell Close from Road ∗ Posti,t + µt + ∈i,t additional 217 TLU (or 217 × 250 = 54,250 kg) of wildlife biomass over a decade compared to cells that remained 50 kilometers from a road. Given that the average wild- Results life biomass in a cell between 1980 and 2009 was 266  TLU, the impact of roads has been significant: It NON-PARAMETRIC EVIDENCE ON ROADS amounts to a 78 percent additional decrease of wildlife. AND WILDLIFE Twenty kilometers from a road, the impact, although two Figure B.2 plots a local smoothing regression (LOWESS) times smaller, remains ecologically significant. between the total wildlife loss in Kenya between the 1980s and the end of the 2000s, and the Euclidean dis- Table B.2 shows the results of the standard parsimoni- tance to the nearest road. Wildlife decreased at a faster ous difference-in-differences model in which results pace closer to roads. It highlights that wildlife loss was remain robust. higher close to roads (5 to 10 km). 100 km from a road, results are no more significant. FIGURE B.2: Distance to roads and wildlife loss 500 250 Wildlife lost (in 1,000 TLU) 0 –250 0 50 100 150 Distance to nearest road (in km) 50  WHEN GOOD CONSERVATION BECOMES GOOD ECONOMICS—Kenya’s Vanishing Herds TABLE B.1: Main model (1) (2) (3) (4) (5) Less than 5 km Less than 10 km Less than 15 km Less than 20 km Less than 50 km Treated × post –217.369* –185.138** –134.558* –114.494* –65.554 (121.325) (85.765) (79.109) (65.091) (44.057) Post –358.628*** –389.410*** –345.288*** –326.846*** –530.919*** (101.365) (100.153) (110.082) (105.581) (143.272) Observations 2,586 2,730 2,868 3,027 4,029 Number of cells 862 910 956 1,009 1,343 Treatment Road becomes Road becomes Road becomes Road becomes Road becomes <5 km <10 km <15 km <20 km <50 km Control Road 50 to 100 km Road 50 to 100 km Road 50 to 100 km Road 50 to 100 km Road 50 to 100 km from cell from cell from cell from cell from cell Note: * = p<0.05, ** = p<0.01, *** = p<0.001. TABLE B.2: Parsimonous model (1) (2) (3) (4) (5) Less than 5 km Less than 10 km Less than 15 km Less than 20 km Less than 50 km Treated × post –207.072* –177.400** –134.719 –122.465* –97.502* (125.676) (88.791) (83.136) (70.353) (56.474) Post –157.666*** –163.580*** –165.305*** –164.635*** –165.006*** (16.259) (16.427) (16.704) (16.633) (16.338) Observations 2,586 2,730 2,868 3,027 4,029 Number of cells 862 910 956 1,009 1,343 Treatment Road becomes Road becomes Road becomes Road becomes Road becomes <5 km <10 km <15 km <20 km <50 km Control Road 50 to 100 km Road 50 to 100 km Road 50 to 100 km Road 50 to 100 km Road 50 to 100 km from cell from cell from cell from cell from cell Note: * = p<0.05, ** = p<0.01, *** = p<0.001. REFERENCE Ogutu, J. O., Piepho, H. P., Said, M. Y., Ojwang, G. O., Njino, L. W., Kifugo, S. C., & Wargute, P. W. (2016). Extreme wildlife declines and concurrent increase in livestock numbers in Kenya: What are the causes? PloS one, 11(9), e0163249. Road Extension and Wildlife Loss between 1980 and 2010: A Difference-in-Differences Approach   51