Lake Chad Regional Economic Memorandum  |  Development for Peace Technical Paper 6. Building Rural Development in the Lake Chad Region Brian Blankespoor (World Bank) 260 Technical Paper 6. Building Rural Development in the Lake Chad Region 7.1 Introduction Limited market accessibility, and more recently Yaoundé. Unpaved roads continue to pose challenges conflict, hinder agricultural production and therefore during the rainy season. pose major challenges to the economic recovery and development of the Lake Chad region. More than Connectivity is very important to the local economy 250 million people live in the four countries of the and cross-border trade. Previous literature demonstrates Lake Chad region, where the vast majority of the people that road infrastructure is conducive to regional trade depend on agricultural activities for their livelihoods. and growth where it can facilitate local economic growth Compounding limited paved roads in agricultural areas with the reduction in input and transportation costs, is the occurrence of numerous violent events from the while connecting the potential for higher prices of crops decade long insurgence of Boko Haram near Lake Chad. (Crawford et al., 2003; Redding and Turner, 2015; Berg Displaced and conflict-affected households face limited et al., 2017; Aggarwal, 2018; Henderson et al., 2017; market access and economic opportunities to earn income Jedwab and Storeygard, 2020; Storeygard, 2016; Jedwab (FEWS NET, 2021). Therefore, it is very timely and and Storeygard, 2019). In Nigeria, roads can facilitate crucial to understand the links between these challenges growth of the non-agricultural sector (Ali et al., 2015) and agricultural activities. and improve access to market that has been linked to the adoption of modern technologies (Damania et al., 2017). The objective of this paper is to gain insight on the Even so, Chamberlin et al. (2014) show that suit able current state of rural development in the Lake Chad land remains uncultivated in Sub-Saharan Africa due countries by examining the relation between road to limited transport access. Using panel methods over investments and cropland expansion  over the past three which significant road development took place from decades following the framework in Berg et al. (2018) 1970 to 2010 in Sub-Saharan Africa, Berg et al. (2018) and then by investigating rural development in proximity demonstrate a modest impact of improved market access to conflict events arisen by the insurgence of Boko Haram on cropland expansion and suggestive evidence of impacts this past decade. on the local intensity of croplands. Over the past several years, roads have improved in Not only are roads crucial to the Lake Chad region, the Lake Chad region  (Magrin et al., 2018), yet limited but these countries are also linked to the water paved roads and road maintenance continues to be a resources of Lake Chad, which is a large area natural challenge. Nigeria has one of the largest road networks transboundary resource that supports local livelihoods in West Africa in terms of length of roads. However, including farming, livestock, and fisheries ( Déby Itno the percentage of national roads in a bad state increased et al., 2015). Due to environmental changes and human from 23 percent in 1985 to 60 percent in 2010 (Federal activities, Lake Chad has shrunk approximately 90 percent Government of Nigeria, 2010). The other three countries from 1960 levels, when it was the world’s sixth largest have less paved roads. The concentration of roads in inland water body. The region is subject to droughts as Niger is mainly in the south along the West-East route well as human activities have altered the hydrology of this of Niamey to Nguime via Diffa as well as the a triangle endorheic lake by stream flow modification and water including Agadez, Tahoua and Zinder. Chad has limited diversion (Lemoalle et al., 2012), which contributes to paved roads, which are mainly concentrated in the capital the water scarcity and fragility of the region (Okpara region. The roads in Cameroon connect the main port of et al., 2015). Droughts can challenge agricultural Douala and the administrative cities including the capital production and correspond to an increase in violence 7.1 Introduction 261 Lake Chad Regional Economic Memorandum  |  Development for Peace against civilians (Bagozzi et al. 2017). In addition to the Although an explicit strategy by Boko Haram to attack administrative challenges of a transboundary resource, markets is not known, markets are still a key location the fluctuations in interannual and seasonal water impede to disrupt trade and target civilians. Using conflict event the development of stable resources exploitation rights data Van Den Hoek (2017) reports 38 direct attacks on (Sarch, 2001) and reduce groundwater discharge along markets in Borno state, Nigeria, between November 2014 with loss in biodiversity (Odada et al., 2003). and December 2016, which is nearly two market attacks per month. The paper also finds seasonality in the timing Along with the environmental changes over the past of the attacks, which occur prior to the lean season and just few decades, the fragility of the region has increased after the harvest. Both of these periods have the potential in the past decade due to the insurgency from Boko to disrupt agricultural production and trade by various Haram in northern Nigeria that contributes to channels: impeding physical access, access to inputs, the humanitarian challenges in the region. Conflict can timing of planting and harvesting, and abandoning of drive population displacement, impede the normal fields. Adelaja and George (2019) examine the effect of activity of local markets, and constrain household access the Boko Haram conflict on agricultural productivity to livelihood, food and income. Conflict events from Boko using a nationally representative panel dataset and Haram started in 2009 have caused massive displacement micro data from the ACLED database.358 They do not of people and disruption to the agricultural sector find a decrease in the total hectares of agricultural land including market activities (Awodola and Oboshi, 2015; harvested, however they do find a significant reduction Van Den Hoek, 2017; Jelilov et al., 2018). The conflict in total output and productivity from the Boko Haram has displaced a large number of individuals who have attacks. Adebisi et al. (2016) find negative impact on experienced significant income shocks with an increase agribusiness in Borno state, Nigeria. Barra et al. (mimeo) of over 40 percent chance of having no income based on examine the relationship between conflict and poverty in an analysis of Nigeria (UNHCR and World Bank, 2016). Nigeria considering the connectivity where they find that From a recent report by OCHA (2020), the Lake Chad decreasing transportation costs with less multidimensional region has 2.6 million Internally Displaced People with poverty. Ali et al. (2015) find that reducing transportation 256 thousand refugees, and 5.2 million people are severely costs in Nigeria increase measures of welfare. food insecure as of 16 September 2020. FAO (2017) recently reported that nearly 50 percent of the 704,000 This paper examines the relationship between access people in Niger are in dire need of humanitarian assistance to markets and land cultivation following Berg et al. and nearly 20 percent are facing issues of food security. (2018) using panel methods. Then, I contextualize these For Nigeria, nearly 70 percent of the 12 million people results within the broader recent development challenges are in need of humanitarian assistance with 43 percent of the Lake Chad region. The results provide evidence facing issues of food insecurity. Employing a Difference that an increase in market access is associated with an in Differences framework with panel household survey increase in cultivated land and is positively associated data, Agwu (2020) finds that exposure of households with an increase in local agricultural GDP. Even so, to conflict events from Boko Haram is associated with conflict from the rise of Boko Haram in the past decade significant downward movements in food security. In a can attenuate gains whereby the proximity to conflict study of Africa, Maystadt et al. (2020) find evidence of events in the previous year is associated with less cropland agricultural expansion near refugee-hosting areas, whereas across the entire region and less night time lights from Salemi (2021) finds evidence of small increases of forest over a hundred local markets nearby Lake Chad. loss (intensive margin and not extensive margin) in areas near refugee camps in Sub Saharan Africa. 358 They use the Living Standard Measurement Study Integrated Study on Agriculture dataset with three waves: 2010–11, 2012–13 and 2015–16. 262 7.1 Introduction Technical Paper 6. Building Rural Development in the Lake Chad Region This paper makes two contributions. First, the The rest of this paper is structured as follows. Section importance of market access as part of economic 2 describes the data sources while section 3 presents the development is well known, yet advancements in empirical framework, section 4 presents the results, and measurement of agricultural activity derived from section 5 concludes. satellite data and recent data are necessary to gain current insight given developments in the region. Following Berg et al. (2018) who examine Sub-Saharan Africa during the period 1970 to 2010, I examine market access for rural development in the Lake Chad countries over a period during which changes in cultivated area and modest road improvements took place. I provide contemporary insights with a higher spatial resolution measure of cropland derived from satellite data at 300m from 1992 to 2019 building on the findings in Berg et al. (2018).359 In addition, I use a newly available data set on agricultural GDP (Blankespoor et al., forthcomingc) to examine the impact of market access on local agricultural GDP.360 Similarly, I examine the local conditions in each grid cell to determine if areas of increased cultivated land are exposed to more suitable agricultural production conditions. I focus on the extensive margin of cropland expansion (rather than intensification) given the strong dependence on rainfed agriculture and the relatively small share of cropland that is classified as irrigated. Second, this paper contextualizes the findings of market access with local conditions given the numerous conflict events in the past decade from Boko Haram. First, I examine the impact of proximity of conflict events on cropland expansion during the period 2009 to 2019 for the entire region. Second, I build on the market level analysis by Van DenHoek (2017) who examines agricultural market activity for 104 markets nearby Lake Chad. I examine the impact of conflict by adding more recent observations of market status in a new framework that includes local night time lights as a proxy for local economic activity and a measure of proximity to Boko Haram events. 359 Before 2000, the HYDE 3.2 database methodology uses a weighting algorithm to estimate cropland including slope where as the measure from ESA is the result of supervised and unsupervised methods using satellite time-series data. 360 Berg et al. (2018) examine the impact of market access on total GDP; they did not examine the impact on the agricultural sector. 7.1 Introduction 263 Lake Chad Regional Economic Memorandum  |  Development for Peace 7.2 Data and Sample Given the challenges of data collection across multiple of agricultural suitability based on soil and climate countries, geospatial methods integrate a variety of data conditions for twelve major crops for the period 1981– at a consistent unit of 0.1 × 0.1 degrees (approximately 2010 from the ClimAfrica project (WP4).362 Another 11km × 11km at the equator) covering the four measure is a common drought index with the Standardized countries surrounding the Lake Chad. I intersect these Precipitation Evapotranspiration Index (SPEI) algorithm grids with the national border to create a total of 33,252 (Beguería et al., 2014) using the monthly precipitation pixels along with the corresponding area. The number and evapotranspiration 1950–2019 data version 4 from of observations depends on the locations and time-step the Climate Research Unit (Harris et al., 2020). I count included in the regression. The regional analyses with full the number of months in a cell that are considered severe geographic coverage includes all pixels times the number drought with values below or equal to -1.5 (Guenang and of years. The geographic definition of the Lake Chad area Kamga, 2014). Given the time scale over which water is from the World Bank Lake Chad Regional Recovery deficits accumulate for agricultural is important and I run and Development (PROLAC) project.361 The local the 18 month lag.363 market level analysis is limited to the pixels with field based observations of a 104 markets summarized at the pixel level by season. These datasets provide insight into 7.2.2 Land cover and agricultural activity agricultural activity given the limited official statistics and access to the field. Below is a description of the datasets. Land cover estimates from satellite data provide a geographically comprehensive and consistent measurement from which to identify trends in 7.2.1 Local conditions agricultural activity ( Weiss et al., 2020). Previous work by Berg et al. (2018) examined cultivated area from 1970 Agricultural production is subject to the local variation to 2010 using cropland estimates in the History Database in annual climate and initial conditions ( Zaveri et al., of the Global Environment (HYDE) 3.2 (Klein Goldewijk 2020). Figure 1 illustrates the shared boundary of Lake et al., 2017). A more recently released dataset is from Chad as well as major transboundary rivers including the the European Space Agency (2017, 2019) that provides Niger and Benue from the Global Lakes and Wetlands annual estimates of land cover at 300m from 1992–2019 Database (Lehner and Döll, 2004). Lake Chad has which are harmonized from two data products Land extensive floodplains and wetlands (Odada et al., 2005). Cover Maps - v2.0.7 and Land Cover Maps - v2.1.1.364 I aggregate the 38 categories into six: irrigated and rainfed I summarize mean annual precipitation and its square cropland, cropland mosaic, grassland, urban, bare land at the cell-year level from the CHIRPS dataset version and other.365 I summarize these data by pixel into the share 2 (Funk et al., 2015). FAO (2013) provides a measure of cropland per pixel.366 I also include another measure of 361 It includes the following administrative areas: Far North Region (Cameroon); Lac, Kanem, Hadjer Lamis, and Chari Baguirmi Regions (Chad); Diffa and Zinder Regions (Niger); and Borno, Adamawa and Yobe States (Nigeria). Notably, the definition excludes N’Djamena in Chad. 362 FAO GAEZ version 4 was unavailable at the time of analysis to measure high agricultural suitability following (Berg et al., 2018). 363 The 18 month SPEI provides information of precipitation patterns over a medium to long time scale. 364 Liu et al. (2018) examine the correlation between cropland area in FAOSTAT and ESA-CCI-LC. 365 Specifically, I aggregate the landcover classes 10–12 into cropland; 20 as irrigated, 30 as mosaic cropland; 150 as grassland, 190 as urban and the remaining codes defined as other. 366 The area considered irrigated is significantly smaller than cropland. 264 7.2 Data and Sample Technical Paper 6. Building Rural Development in the Lake Chad Region cropland that includes more than 50 percent mosaic crop measures below at the season level. Another measure to lands with less than 50 percent natural vegetation (tree, gain insight on agricultural activity is from a measure of shrub, herbaceous cover) to include mixed land use with burned areas derived from satellite. Burning agricultural smallholder agriculture. fields is a common practice in Central and West Africa and can reduce post-harvest fires as well as provide short- The land cover model provides an estimate of annual term nutrients (Bucini and Lambin, 2002; Kull and Laris, agricultural land use, however it aggregates any 2009). I summarize the area from the MODIS Burned seasonal variation. So, following Blankespoor et al. Area data product (v6) in a cell, which provides a burned- (forthcomingb) and a calendar with the seasons of major area estimate per 500m pixel by month (NASA, 2020a). crops for each country from FAO (2020a,b,c,d) I define The intensity of land use is measured from a greenness each month into three seasons: (i) land preparation; (ii) measurement called Normalized Difference Vegetation sowing and growing; and (iii) harvest. I summarize the Index (NASA, 2020b). Another measure of vegetation  his map illustrates the distribution and type of waterbodies from Lehner and Döll Map 7.1: T (2004) and the Lake Chad area within the solid black line 7.2 Data and Sample 265 Lake Chad Regional Economic Memorandum  |  Development for Peace growth is Net Primary Productivity that captures on the road types and topography between locations i the solar energy absorbed by plants or other primary and j at time t through the values τi,j–σ and excluding producers (Running et al., 2004). Previous work shows itself and cells within 20km. Following previous studies a strong positive correlation between these estimates and (Berg et al., 2018; Jedwab and Storeygard, 2020), I use crop yields (e.g. Strobl and Strobl, 2011; Zaveri et al., the value for the elasticity of trade σ equal to 3.8 from 2018). The source is from the MODIS satellite product Donaldson (2018) who derived it for the case of India. (MOD17A2H) as summarized by monthly mean of a The travel time is calculated based on a time cost raster cumulative 8-day composite with a 500m resolution method using the Dijkstra algorithm as the minimum (NASA, 2020c). Cross-sectional spatial distribution of time result from the roads and offroad speeds. For the crops such as cotton are from the Spatial Production years 1983, 1992, 2001 and 2010, I assign a speed based Allocation Model (SPAM) (Yu et al., 2020). on road categories similar to Berg et al. (2018); Jedwab and Storeygard (2020)368 and offroad speed based on the In addition to cropland, the livestock and fishing hiking function from (Tobler, 1993). trade are important activities in the Lake Chad region, especially near Lake Chad. The spatial distribution of Hiking = 6 * e–3.5*|s+0.05| * 0.6 livestock ca. 2010 is from Gilbert et al. (2018). They provide estimates livestock including cattle and goat where s is mean slope from Verdin et al. (2007).  based on agricultural census data with equal weights.367 Lake Chad was once one of the great fisheries. Graaf et For panel roads, I use the georeferenced panel roads al. (2014) in FAO (2017) estimate of fishing activities data from 1983 to 2010 from(Jedwab and Storeygard, in the region at a value of USD 54 to 220 million. 2019). The length of paved roads increased in Chad However, current indiscriminate fishing practices are not along with Cameroon, while Niger and Nigeria does not sustainable, yet employed as a coping strategy for survival increase significantly (See Map 7.2). (Eriegha et al., 2019). The size of the market is based on urban population data from the consolidated urban population database 7.2.3 Market Access (Blankespoor et al., 2017). The population has increased over the past decades. The distribution of population in Following Jedwab and Storeygard (2019) and Berg the Lake Chad area has high variation, where Nigeria has et al. (2018), I calculate the local market access for a much higher population than the other three countries. given location as a function of the weighted sum of the populations of all other locations, with a weight that The regional analysis also considers other important decreases with travel time. Formally, I define market infrastructure and markets. Ports provide an important access in a location i at time t: connection to the international market. The location of marine ports is from the World Ports Index (National M Ai,t = ∑j≠i Pj,t τi,j–σ   (1) Geospatial-Intelligence Agency, 2014). The busiest marine port in Cameroon is Douala followed by Limbé.  here Pj,t is the population in location j at time t, τi,j is w Nigeria has major ports including: Lagos, Calabar, Onne, the travel time between locations i and j at time t, and Port Harcourt and Warri. I construct a variable estimating σ is a trade elasticity parameter. Market access depends the minimum travel time to a port. The locations of 367 They also provide a version as the result of statistical models with dasymetric weighting. 368 Specifically, highway speed is 80kph; paved is 60kph; improved is 40kph and earthen is 12kph. 266 7.2 Data and Sample Technical Paper 6. Building Rural Development in the Lake Chad Region  ap illustrating highway (red), paved (black) and improved (pink) roads from (Jedwab and Storeygard, Map 7.2: M 2020) cotton ginning factories are digitized from a map on below status into a combined below normal category for a Cotton Zones, Ginning Factories and Exports of West Africa total of three categories. They produced this report every in OECD (2006). The locations of regional livestock few months starting with the earliest publicly available markets are from FEWS NET (2009) for Chad and market activity report in January 2015, which focused Niger FEWS NET (2008) for Nigeria and from Motta et on market activities in December 2014. I summarize the al. (2019) for Cameroon. Within the Lake Chad region, most restrictive category during the season and I exclude the trade routes depend on connectivity of infrastructure 32 missing observations during the six year period of across borders, especially Chadian livestock along with record (N = 104 * 3 * 6 - 32 = 1840); including a one the collection of Cameroonian livestock on the way to year lagged variable reduces the number of observations Nigeria (Magrin et al., 2018). equal to 1533. For local markets near Lake Chad, the Famine Early Warning Systems Network (FEWS NET) reports 7.2.4 Economic activity provide both time and place of the operational status, which are based on field-based investigations, into Night time lights can proxy total local economic activity four qualitative categories:(i) normal activity/operating (Henderson et al., 2011) and human development normally (ii) some disruption, reduced activity/operating (Bruederle and Hodler, 2018). I use two night time lights slightly below normal, (iii) significant disruption, limited datasets due to the time-step. For annual trends across the activity/operating well below normal or (iv) minimal region from 1992–2018, I use the dataset that harmonizes or no activity/not operating. Following Van Den Hoek two data sources DMSP-OLS and VIIRS data from Li (2017) I use the closed and normal operational status et al. (2020).369 The second source is the monthly data and then I aggregate the well below normal status and from VIIRS available from April 2012 until 2020.370 369 DMSP-OLS (1992–2013) is used for temporal calibration with a simulation of VIIRS data (2014–2018). These satellites measure light at night at different times of day. 370 I calculate these values using the Stray Light Corrected Nighttime Day/Night Band Composites Version 1 product. 7.2 Data and Sample 267 Lake Chad Regional Economic Memorandum  |  Development for Peace After visual inspection and inline with previous work by data fusion method based on cross-entropy optimization Li et al. (2020), I calculate the sum of the radiance values that disaggregates administrative level agricultural GDP above 0.3 by pixel.371 Even with the striking correlation of into grids depending on satellite-derived indicators of night time lights and total GDP, these measures require the components that make up agricultural GDP, namely areas to emit light at night to relate to economic activity, crop, livestock, fishery, hunting and timber production. which is not prevalent in many rural areas (Thomas et Map 7.3 illustrates the distribution of agricultural GDP al., 2019). So, it does not account well for a significant circa 2010. The cropland component takes advantage of contribution to the economy from the agricultural sector. the SPAM model to inform the prior allocation of cropland Over the past two decades, Chad, Niger and Nigeria production value and does not directly use infrastructure have a higher share of agricultural GDP than the Sub- data (Yu et al., 2020). The level of agricultural GDP in Saharan Africa regional aggregate.372 To fill this local Nigeria is considerably higher compared to Cameroon, data gap, (Blankespoor et al., forthcomingc) employ a Chad and Niger.  his map illustrates main roads along with the distribution of Agricultural GDP (2010) Map 7.3: T from (Blankespoor et al., forthcomingc), where darker red represents relatively higher agricultural GDP and light blue or transparent has little estimated value 371 Values less than 0.3, which include negative values, are considered background noise such as large areas of the Sahara desert. 372 The World Bank World Development Indicators reports that the share of agricultural added-value GDP is in a range of 15–42 percent. 268 7.2 Data and Sample Technical Paper 6. Building Rural Development in the Lake Chad Region Livelihood zones for each country are from FEWS. 7.5).374 The ACLED database includes over 4,800 events NET. These geographic zones group people with similar with more than 35,000 fatalities that are associated patterns considering how people gain access to food with Boko Haram as an actor from 2009 to December and income as well as markets. The map data include 2020. Many conflict events are in close proximity to the following number of zones: 36 in Nigeria (FEWS Maiduguri, which is the state capital of Borno and major NET, 2018), 9 in Chad, (FEWS NET, 2011a), 15 commercial center in the Lake Chad region. According in Niger (FEWS NET, 2011b) and 17 in Cameroon to a news source in 2013, approximately 5,000 hectares (FEWS NET, 2019a).373 I aggregate these categories into of agricultural plots with wheat and rice were abandoned 29 categories based on the first or dominant crop listed in near Marte in Borno state, Nigeria. This translated to the description with multiple crops (See Map 7.4). roughly 200 metric tonnes of wheat according to Abubakar Gabra Iliya, head of the Lake Chad Basin Development Agency.375 After 2013, conflict events continued to 7.2.5 Conflict data increase in occurrence and spread to include the area across the Nigerian border in the Lake Chad area (See The insurgency by Boko Haram in the Lake Chad area Map 7.5). In Niger, Boko Haram activities target Diffa, has led to an increase in the number of conflict events Bosso and the small villages along the river Komadougou. and fatalities since 2009 with a notable concentration In Chad, Boko Haram is present in the islands of Lake in the three states of Northeastern Nigeria (See Map Chad as well as attacks in N’Djamena, Guitté, Bo and  his map illustrates the livelihoods of the four countries, which clusters similar Map 7.4: T livelihood patterns into a zone 373 The zones for Cameroon were digitized given the lack of response of FEWS NET to provide the georeferenced data and may include small digitizing errors. 374 Violence from Boko Haram increased most notably after the execution of its leader Mohammed Yusuf in July 2009. 375 https://www.pmnewsnigeria.com/2014/03/25/food-supply-crisis-imminent-in-nigeria/ 7.2 Data and Sample 269 Lake Chad Regional Economic Memorandum  |  Development for Peace  hese maps in the panel show the evolution of the number of events from 2009–2020 defined as Boko Map 7.5: T Haram Source: ACLED (Raleigh et al., 2010) (downloaded 2020-05) and author’s calculations. Baga Sola. More recently, Boko Haram has taken refuge in the Sambisa forest, which is South East of Maiduguri, and the swamps of Lake Chad (Magrin et al., 2018). Conflict has taken place, especially in the area in close proximity to Lake Chad. I use the location of conflict events from the Armed Conflict Location & Event Data Project (ACLED) (Raleigh et al., 2010). I summarize the number of events and fatalities by cell. 270 7.2 Data and Sample Technical Paper 6. Building Rural Development in the Lake Chad Region 7.3 Empirical Framework In a framework following Berg et al. (2018), I first to the international market. To account for heterogeneous explore the links between market access and two effects, I include an interaction of the natural logarithm measure of agriculture: cropland area and a local measure of the market access index with a measure of agricultural of agricultural economic activity. I examine the impact of suitability and a measure of shrinking land. The market access on cropland area using a panel framework regressions include country fixed effects, time dummies, and a second model exploring the association between and the interaction between the two as a control for any cropland area and local agricultural activity (Agricultural remaining unobserved heterogeneity. GDP). Given numerous conflict events have occurred in the past decade, I also examine agriculture activity amidst conflict in a panel framework examining the association 7.3.2 Local agricultural activity across of the proximity of conflict on cropland area across the the region entire region as well as night time lights for a sample of local markets nearby Lake Chad. I also explore the association between cropland area and a measure of local agricultural activity (local agricultural GDP) at the grid cell level w  ith the 7.3.1 Cropland expansion across the following regression estimated in levels. The regression is region defined as follows: The cropland regression in levels is defined as follows:  lnAgGDPi,t = β0lnCropi,t–9 + β1lnM Ai,t–9 + X'i,t–90 + D'iπ + δt + ϵit  (4) InCropi,t = α0InM Ai,t–9 + X'i,t–9θ  here InAgGDPi,t is natural logarithm of the local w + D'iπ + δt + ϵi,t  (3) agricultural GDP in pixel i for time period t (2010),  here InCropi,t is natural logarithm of the area of cropland w InCropi,t–9 is natural logarithm of the area of cropland in pixel i for time period t, InMAi,t–9 is the lagged natural in the previous period of 9 years, InMAi,t–9 is the lagged logarithm of the market access indicator is the result of natural logarithm of the market access indicator is the Equation (1), Xi,t–9 is a vector of control variables at time result of Equation (1), Xi,t–9 is a vector of control variables t, D'i is a vector of time invariant dummies, and ϵi,t is the at time t, D'i is a vector of time invariant dummies, and error term. ϵi,t is the error term. Using the same approach to address concerns of reverse The other controls are similar to Equation (3). The causality I employ a lag in the market access index travel time to livestock markets accounts for proximity by one period (9 years) as cultivation may influence to livestock trading. The regressions include country the placement of new road investments as well as the dummies given the agricultural GDP dataset is only changes in local population. To account for the local level available for one time step. A cautionary note is the local of population, I include population density estimated by agricultural GDP is the result of a cross-entropy model UNEP-GRID Geneva and The World Bank. The travel that leverages spatial detail in the subcomponents of time to the nearest major port is a measure of proximity agricultural GDP.376 376 See Thomas et al. (2019) for more details and model comparisons. 7.3 Empirical Framework 271 Lake Chad Regional Economic Memorandum  |  Development for Peace 7.3.3 Agriculture amidst conflict nearby the market operational status on the natural logarithm Lake Chad of night time lights. Then, I introduce a distance to the nearest conflict event from the previous year into the The previous analyses examine the impact of market regressions as follows: access over the past three decades; this framework does not account for the lived reality of access to markets lnNTLi,t + α0Marketi,t + α1distConflicti,t–3 + X'i,t–1θ + D'iπ + δt + ϵi,t  (6) given the insurgency of violence on an annual or seasonal basis. The next section focuses on examining  here InNTLi,t, is natural logarithm of the sum of night w agricultural activity amidst conflict. time lights in pixel i for time period t, which is defined as a season of the year (land preparation, sowing and growing, or harvest). Market is the operational status 7.3.3.1 Cropland expansion amidst conflict of the local market as normal, below normal or closed. distConflict is the distance to the nearest conflict event in  examine the association of distance to nearest conflict I pixel i during the same season of the previous year, Xi,t is a event or fatality on local cropland extent during the vector of control variables at time t, D'i is a vector of time period 2009 to 2019 for the entire region. invariant dummies, and ϵi,t is the error term. I include the mean precipitation and its square along with mean lnCropi,t = α0lnDistConfi,t–1 + lnDistConfi,t–1 greenness during the season as a control, which is used as × Yi + lnNTLi,t–1 + lnNTLi,t–1 × Si,t–1 + lnNTLi,t–1 × lnM Ai,2008 + X'i,t–1θ a proxy for local agricultural productivity. + X'i,t–10 +D'iπ + δt + ϵi,t  (5)  here lnCropi,t, is the natural logarithm of the cropland w area in pixel i for time period t in years, lnDistConf is the natural logarithm of the nearest distance to a conflict event or an event with a fatality from pixel i in the previous year t – 1, Xi,t–9 is a vector of control variables at time t, D'i is a vector of time invariant dummies, and ϵi,t is the error term. I use a lag in the conflict variables to address reverse casualty. I address the concern about modeling error of local population estimates due to displacement by taking advantage of the annual frequency and high correlation of night time lights with population density. I include lagged night time lights variables to control for size effects. Then, I interact night time lights with shrinking cropland, market access in the previous base year (2008) and natural logarithm of the travel time to nearest livestock market in 2008. 7.3.3.2 Operational status of markets amidst conflict For a set of markets in the area nearby Lake Chad with reported operational status, I examine the impact of 272 7.3 Empirical Framework Technical Paper 6. Building Rural Development in the Lake Chad Region 7.4 Results 7.4.1 Trends in Agriculture sorghum as well as irrigated crops such as Komadougou Irrigated Peppers and Violet de Galmi onions (FEWS Agriculture is the main sector of economic activity NET, 2011b). Northern Nigeria has cultivated areas for households and individuals living in the Lake with diverse crops including millet and sorghum as well Chad region. For the four countries, Map 7.4 illustrates as livestock. The area nearby Lake Chad includes flood dominant livelihoods with similar patterns considering retreat cultivation and fishing activities. Major cash crops how people gain access to food and income as well as include cotton in the Sahel region where many cotton markets. The northern areas of Niger and Chad are ginn factories are located (See Map 7.6). Other rainfed sparsely populated with activities including salt, dates and cash crops such as bananas and coconuts are located near trading activities in oases along with nomad pastoralism the ports outside of the Lake Chad area. Maize, cassava, and transhumance. In southern Niger, where most of the sorghum and millet provide staple crops and are among population lives, is an agropastoral belt with millet and the highest production in the Lake Chad region.377  his map illustrates cotton production from SPAM ca. 2010 (Yu et al., Map 7.6: T 2020) and ginning factories (OECD, 2006) 377 These results are summarized from the SPAM circa 2017. 7.4 Results 273 Lake Chad Regional Economic Memorandum  |  Development for Peace In the region, agricultural production is typically rain- the lesser intensity one in the 00s, where the Lake Chad fed and thus dependent on the climate. The levels of area typically has a higher share of drought compared to rainfall have varied over the past several decades. One can the remaining are in the country. However, the period consider three periods of climate since 1960 in the Lake from the 1980s until present has seen some greenness Chad area: a period of high rainfall in the 1960’s, low growth as measured from satellite (e.g. Dardel et al., rainfall in 70s to the 90s and the recent period with more 2014). As shown above, some seasons drought persists, variability than the two previous periods. Droughts can however the opposite case is also true. In 2019, flooding impact food availability and timing and a households’ occurred in an area with approximately 220,000 people ability to consume. Figure 7.1 displays the share of area as the result of heavier-than average rainfall in the fall that exceeds a drought threshold, which is measured by of 2019. USAID (2020) reported the damage from the SPEI with 18 month lag, for the Lake Chad study region floods included infrastructure, crops and restricted access. "inside" and the remaining area in the country "outside". The area nearby Lake Chad is rural and dependent on One can see the major drought in the 70s and 80s378 and hydro-climatic conditions (Nilsson et al., 2016).  his graph shows the share of area considered a drought identified from the Standardized Figure 7.1: T Precipitation Evapotranspiration Index (SPEI) at an 18 month time lag (with a value less than or equal to -1.5). This is the result of the algorithm provided by Beguería et al. (2014) using the monthly precipitation and evapotranspiration data version 4 from the Climate Research Unit (Harris et al., 2020) 378 Rivers did not flow into the lake in 1973 and 1984 as the result of Sahelian drought (Raji, 1993). 274 7.4 Results Technical Paper 6. Building Rural Development in the Lake Chad Region  hese maps illustrate the location of livestock markets (black dots) from FEWS NET (2009); Motta et al. Map 7.7: T (2017) and distribution of cattle from Gilbert et al. (2018) (left) and goat (right) in 2010 from Gilbert et al. (2018) where a darker shade represents higher livestock density  his map illustrates land use (Nachtergaele et al., 2010) (Version 1.1) Map 7.8: T 7.4 Results 275 Lake Chad Regional Economic Memorandum  |  Development for Peace Over the past two decades, arable or cropland areas Harcourt-Enugu (See Map 7.3). The natural logarithm increased although irrigated lands are limited as of the distance to nearest paved road in 2008, which is measured from satellite imagery. Total cropland area constrained by country, has a negative correlation with has increased by almost 43,000 km2 between 1992– natural logarithm of local agricultural GDP of 0.70. 2019, where Cameroon and Nigeria have the most relative gain in cropland area compared to Chad and In closer proximity to Lake Chad, road transport Niger. Irrigated areas represent approximately 5 percent connects key local agricultural markets  of (i) Bol of the cropland area during this period with little growth, and N’Djamena in Chad, (ii) Kousseri in Cameroon, which is the likely result of little new investment. The (iii) N’guigmi and Diffa in Niger and (iv) Bosso, Niger annual growth rate of cropland area for the four countries along with Marte and Monguno via Madiguri in Nigeria. started at 0.42 percent during the 1992–2001, lessened at Maiduguri is an important connection for the trade 0.31 percent during 2001–2010, and was lower at 0.03 corridors between Nigeria and Cameroon (Kousseri or during 2010–2019.The annual growth rate of cropland Maroua) (See Map 7.9). This was especially important area for the four countries inside the area nearby Lake in the commercialisation of fishing in Lake Chad linked Chad started at 0.29 percent during the 1992–2001, to the development of road infrastructure (Stauch, A., lessened at 0.24 percent during 2001–2010, and was 1960). For the fishing trade, a number of fish markets slightly negative during 2010–2019. exist; the largest fish market was Baga Kawa in Nigeria (prior to Boko Haram), which is a key market town in In addition to cropland, the livestock trade is vital for close proximity of Lake Chad (Magrin et al., 2018). the region and cross-border trade has long played a role in trade in livestock markets in Africa  (de Haan  his map illustrates local markets in close Map 7.9: T proximity to Lake Chad by type et al., 1999). The distribution of livestock markets and a subnational estimate of cattle is notable in the Lake Chad area (See Map 7.7). Over half of the livestock markets are within 100km of the border and 16 of the 97 livestock markets are located in the Lake Chad area. Nearby Lake Chad, agricultural activities benefit from the connection to markets. Fish routes supply several tonnes of fish everyday to regional hubs of N’Djamena and Maiduguri with an annual estimate of 50,000 to 100,000 tonnes of fish per year (Lemoalle and Abdullahi, eds, 2017). Similarly, livestock trade routes from Chad and Niger pass through Maiduguri onto regional markets and are an important part of the goat and millet trade (WFP, 2016). Mapping the cropland, livestock, forestry, fishing across the landscape illustrates a mosaic of agricultural activity and management of land  (See Map 7.8). The agricultural economy of Nigeria is notably higher than the other three countries. The development of highways Source: Déby Itno et al. (2015). in Nigeria coincides with the area with relatively higher agricultural GDP circa 2010; they are the three corridors of Kaduna-Kano (North Central), Lagos-Benin City (South West) and the delta region corridor of Port 276 7.4 Results Technical Paper 6. Building Rural Development in the Lake Chad Region 7.4.2 Market Access market access is associated with a 3.9 per cent increase in cropland area. Given the approximate total of cropland Table 7.1 presents the panel results of market access in the four countries is nearly 600,000 km2, this result on cropland area (in levels). The baseline estimation implies a growth of around 23,400 km2 given a 1 percent of equation (3) of market access does have a significant increase in market access over 9 years. Following Berg positive effect on an increase in cropland across OLS et al. (2018), I examine spatial heterogeneity with (Columns 1–3) and FE specifications (Columns 4–6) interactions of market access with yield and a shrinking during the period from 1992 to 2019.379 cropland dummy. The sign of the yield interaction is positive, providing a positive association of the growth The results from new and updated measurements with a in cropland where the land has higher yield of cotton.380 focus on the Lake Chad region are in line with previous Also, it is important to note that the area near Lake Chad research by Berg et al. (2018). Given the modest gain in is landlocked. The results show a connection of cropland length of paved road, the growth in population, which area to external markets measured in travel time to the is a proxy for the size of the market, is the main driver nearest port that are located in Nigeria and Cameroon. for the increase in market access. A 1 percent increase in  stimates of the impact of market access on cropland area Table 7.1: E (1) (2) (3) (4) (5) (6) 0.317*** 0.0420** 0.0374 0.274*** 0.0385* 0.0434* Ln MAt–9 (18.96) (2.72) (1.80) (16.50) (2.39) (1.99) 0.0701** 0.0630* Ln MAt–9 × Yield (2.71) (2.31) -0.0920*** -0.0841*** -0.0826*** -0.0706*** Shrinking (-14.04) (-11.65) (-12.65) (-9.83) -0.0322* -0.0486*** Ln MAt–9 × Shrinking (-2.43) (-3.58) -0.109*** -0.107*** -0.0355*** -0.0332*** Ln time to portt–9 (-14.97) (-14.75) (-4.40) (-4.09) 0.359*** 0.359*** 0.135*** 0.133*** Ln pop densityt–9 (97.01) (97.03) (36.73) (36.34) -0.114*** -0.111*** -0.187*** -0.182*** Mean Precipitationt (-6.33) (-6.16) (-10.42) (-10.21) -0.0119* -0.0114* 0.0269*** 0.0274*** (Mean precipitationt)2 (-2.22) (-2.13) (5.12) (5.18) Country x Year dummies Y Y Y Y Y Y Observations 133,008 133,008 133,008 133,008 133,008 133,008 R-squared 0.0528 0.532 0.533 0.0528 0.449 0.450  statistics in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001 t Notes: This table presents estimates of OLS (column 1-3) and FE (column 4-6) regressions of the natural logarithm of the sum of rainfed and irrigated cropland area at time t on the natural logarithm of the lagged market access index during the period between 1992 and 2019. The controls included in the OLS regressions (column 3-4 and 5-6) include a dummy variable indicating a decrease in cropland during the previous period (shrinkingt-9), the lagged natural logarithm of time to nearest major port (Ln time to major portt-9), the average rainfall over the previous five years / 1000 (Mean precipitationt) and its square, and country × year dummies. Constants are not shown. 379 Specifically, the panel includes the following years: 1983, 1992, 2001, 2010 and 2019. 380 The yield of cotton is a constant value circa 2010 from the SPAM model. Regressions (not shown) of market access on cropland provide similar results with a positive and significant interaction with the agricultural suitability. 7.4 Results 277 Lake Chad Regional Economic Memorandum  |  Development for Peace A 1 percent decrease in this time is associated with a without market access. I find a positive and significant 3.6 percent increase in cropland area. coefficient for market access even beyond the effect of cropland expansion controlling for time to port and time Next, Table 7.2 presents the results examining to nearest livestock market. the association of cropland expansion on local agricultural GDP from the Equation (4). As mentioned These two analyses provide suggestive evidence for earlier, this result is a descriptive association due to the positive impact of market access on cropland potential endogeneity concerns about the modeling expansion and local agricultural economic activity of the local agricultural GDP measure. Cropland does in the countries comprised of the Lake Chad region. have a significant positive effect on an increase in local Given the conceptual framework, these results used a agricultural GDP in OLS specifications.381 Following lagged approach with a period of 9 years. Remarkably, the Berg et al. (2018), I examine the relationship with and current development status in the region has changed since the onset of Boko Haram in 2009. In the next section, I  stimates of the impact of market access on Table 7.2: E examine the association of the location of conflict events Agricultural GDP on cropland and the association of operational status of (1) (2) (3) markets on night time lights. 0.950*** 0.342*** 0.342*** Ln cropt–9 (245.43) (71.44) (71.50) 0.0872*** 7.4.3 Market status and conflict nearby Ln MAt–9 Lake Chad (3.92) -0.312*** -0.302*** Ln time to portt–9 Recent developments with the variation in (-21.67) (-20.46) environmental conditions and conflict pose challenges Ln time to -0.0232*** -0.0242*** livestock for agricultural activity. From the suggestive evidence marketst–9 (-3.27) (-3.41) above, market access is associated with an increase in Ln population 0.316*** 0.315*** cropland area (extensive margin), however this result does densityt–9 (41.44) (41.19) not incorporate short-term shocks or the uncertainty to 1.234*** 1.245*** travel to market, especially related to the proximity of Mean precipitationt (37.78) conflict events.382 The discussion below focuses on the (37.74) Boko Haram conflict regionally and then geographically (Mean -0.259*** -0.261*** on an area in close proximity to Lake Chad with more precipitationt)2 (-27.54) (-27.54) detailed data (e.g. market data) to contextualize current Country x Year development with remotely sensed measures. Y Y Y dummies Observations 33,252 33,252 33,252 R-squared 0.602 0.878 0.878 7.4.3.1 Cropland amidst conflict  statistics in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001 t Notes: This table presents estimates of OLS (column 1–3) regressions of the natural logarithm of Agricultural GDP (ca. 2010) on the natural logarithm of the sum of rainfed and irrigated cropland area (Ln croplandt-9) and the lagged market access index (Ln MAt-9). The As stated above, the Boko Haram insurgency started controls included in the OLS regressions include the lagged natural logarithm of time to nearest major port (Ln time to major portt-9), the lagged natural logarithm of time to nearest livestock market (Ln time to livestockt-9), the lagged natural logarithm of time to nearest near Maiduguri in 2009(see Map 7.5). Numerous reports ginn factory (Ln time to ginn factoryt-9), the average rainfall over the previous 5 years * 1000 (Mean precipitationt) and its square, population density (Ln pop densityt-9), and country × year state the destruction of cropland and infrastructure as dummies. Constants are not shown. 381 The local agricultural GDP data are only available circa 2010. 382 The travel time assumes fastest route and does not include any measures of delays or road blocks. For example, Van Der Weide et al. (2018) incorporate road closure obstacles in the travel time analysis to quantify the impact of market access on local GDP in the West Bank. 278 7.4 Results Technical Paper 6. Building Rural Development in the Lake Chad Region  stimates of the impact of the proximity of conflict on cropland area Table 7.3: E (1) (2) (3) (4) 0.0154*** 0.0176*** 0.0139*** 0.0163*** Ln NTLt–1 (4.50) (5.12) (4.04) (4.73) 0.00413*** 0.00576*** Ln dist to eventt–1 (9.60) (13.38) 0.00237*** 0.00162*** Ln dist to eventt–1 × Yield (17.26) (12.04) -0.0223*** -0.0222*** -0.0225*** -0.0223*** Ln NTLt–1 × Shrinkingt–1 (-15.70) (-15.63) (-15.77) (-15.71) 0.00711* 0.00721* 0.00686* 0.00688* Ln NTLt–1 × Ln MA2008 (2.50) (2.54) (2.40) (2.42) -0.00250*** -0.00289*** -0.00223*** -0.00262*** Ln NTLt–1 × Ln time to livestock2008 (-4.06) (-4.68) (-3.62) (-4.25) 0.0126* 0.0182*** -0.0127** -0.00844 Mean precipitationt (2.55) (3.66) (-2.65) (-1.75) -0.00551*** -0.00724*** 0.00113 0.000184 (Mean precipitationt)2 (-3.54) (-4.62) (0.74) (0.12) Country x Year dummies Y Y Y Y Observations 365772 365772 365772 365772 R-Squared 0.326 0.327 0.180 0.199 t statistics in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001 Notes: This table presents estimates of OLS (1-2) and FE (3-4) regressions of the natural logarithm of cropland area at time t (in years) on the natural logarithm of the distance to the nearest conflict event (Ln dist to eventt-1), night time lights (Ln NTLt-1), the average rainfall (Mean precipitationt) and its square. Constants are not shown.  his panel set of maps shows the evolution of cropland for three distinct periods: 1992–2000 (left Map 7.10: T map); 2000–2009 (center map); 2009–2018 (right map) Source: ESA. 7.4 Results 279 Lake Chad Regional Economic Memorandum  |  Development for Peace  wo figures illustrating the seasonality of the number of conflict events with NDVI (a) and NPP (b) Figure 7.2: T  his graph displays the monthly frequency of (a)  T  his graph displays the monthly frequency of (b) T conflict events from Boko Haram and the mean conflict events from Boko Haram and the mean level of greenness of vegetation, as measured by level of Net Primary Productivity, as measured by NDVI. NPP. Sources: Raleigh et al. (2010); NASA (2020b) and author’s calculations. Sources: Raleigh et al. (2010); NASA (2020c) and author’s calculations. well as the undermining the supply routes of agricultural Recent research finds that the rise of Boko Haram inputs (e.g. FAO, 2017; Jelilov et al., 2018). I find results in more agricultural burning  ( Jedwab et al., suggestive empirical evidence of the decrease in cropland forthcoming), which has been associated with agricultural as derived from satellite land use classification in areas activity due to the common practice of burning fields near conflict events. During the period 2009 to 2018, for clearing and (short-term) nutrients (Blankespoor cropland area in North East Nigeria decreased, even et al., forthcomingb). Map 7.11 shows the change in though the previous two periods indicated some growth burning activity since 2001 in four time periods. The in cropland area (See Map 7.10). Table 7.3 presents the first period shows the variation in burned areas before results of the OLS and FE regression of the impact of the Boko Haram, whereas the second panel map illustrates proximity of conflict from Boko Haram on cropland area a reduction in burned density. Notably, the third panel from 2009 to 2019. The results highlight an association shows a concentration of burned area nearby Maiduguri of an increase of cropland area away from the natural and Dikwa in Nigeria that may be attributable to both logarithm distance to the nearest event or fatality during clearing of fields and conflict events (e.g. the burning the previous year. of buildings), whereas Northern Cameroon indicates reduction in activity relative to 2013. The last panel of Given the seasonality of agricultural production, data (2017–2019) shows an attenuation of these burned Figure 7.2 shows measures of crop phenology along areas. So, it is important to note that this measure can with the frequency of Boko Haram conflict events. capture both agricultural activity and conflict, so it is Focusing on the three Northeastern states in Nigeria, necessary to examine the description of the conflict these events have two peaks: one during the harvest and events.383 another in the land preparation stage. 383 The ACLED database include a description of conflict events that note burning of buildings or razing village(s). 280 7.4 Results Technical Paper 6. Building Rural Development in the Lake Chad Region  his panel map illustrates the distribution of burned density during the harvest season Map 7.11: T Source : MODIS. 7.4.3.2 Markets amidst conflict recommendations for some markets to close given the sites are targets for attacks (on civilians) (FEWS NET, Previous analysis associated the impact of Boko Haram 2015). Markets in the Diffa region were officially shut on the operational status of local markets  (Van Den down to impede supply routes to the insurgents and Hoek, 2017). The spatial concentration of Boko Haram markets in the Far North of Cameroon closed in response events was primarily in Northeast Nigeria. Although the to repeated suicide bomber attacks (FAO, 2017). From monitoring of the markets from FEWS NET started only 2017–2020, several markets on the fringe operated with a in 2014, conflict events already took place by Boko Haram slightly below or normal status. Notably, markets in close at local markets in Nigeria, especially in Borno state proximity to Lake Chad were well below or not operating. (Awodola and Oboshi, 2015).384 Map 7.12 illustrates the More recently in 2020, markets in Chad near the border evolution of market status using market data from 2014– with Cameroon and Nigeria were not operating. Food 2020 based on FEWS NET in Van Den Hoek (2017) and Agriculture Organization of the United Nations with additional digitized time periods from FEWS NET (FAO) reports that the conflict situation continue to pose (2019b, 2020).385 Before 2014, nearly all of the conflict challenges for household to access land and agricultural events from Boko Haram took place in Nigeria (Raleigh inputs. et al., 2010; Jedwab et al., forthcoming). Many markets were not operating from 2014 to 2016 despite regained Table 7.4 presents the panel results of operational territory from some recovery efforts in 2015 from West status on the natural logarithm of the mean of monthly African troops. Physical damage also took place to market night time lights during the season of the year. The infrastructure, for example, over 650 shops were reported baseline estimation of equation (6) of the operational as damaged in Damaturu, Yobe, Nigeria (Mercy Corps market status compared to the reference closed market et al., 2017). In addition to the indirect and direct does have a significant positive effect on an increase in impacts of Boko Haram, the Nigerian government made night time lights for both normal and below normal 384 ACLED database has 23 events with "market" in the notes between 2012 and 2013. 385 The first available year is 2014 along with updates at irregular intervals. 7.4 Results 281 Lake Chad Regional Economic Memorandum  |  Development for Peace  his panel set of maps shows the evolution of market status in and near Northeast Nigeria with a Map 7.12: T selection for each year from 2014 to present with the month that the report was published Source : Van Den Hoek (2017); FEWS NET (2019b, 2020) and author’s calculations.  stimates of the operational status of local markets on night time lights Table 7.4: E (1) (2) (3) (4) 0.0122** 0.00699 0.0117** 0.00670 Market : belowt (0.00484) (0.00487) (0.00483) (0.00472) 0.0182** 0.0155** 0.0171** 0.0147* Market : normalt (0.00799) (0.00727) (0.00825) (0.00757) 0.00784* 0.00781* Ln dist. to conflict eventt–3 (0.00418) (0.00451) 0.170*** 0.195*** 0.176*** 0.202*** Mean NDVIt (0.0308) (0.0331) (0.0318) (0.0332) -0.531*** -0.677*** -0.532*** -0.676*** Precipitationt (0.0870) (0.107) (0.0911) (0.111) 0.662 1.224* 0.619 1.149 (Mean precipitationt)2 (0.643) (0.673) (0.674) (0.713) Country x Year dummies Y Y Y Y Observations 1,840 1,533 1,840 1,533 R-Squared 0.189 0.180 0.189 0.181  tandard errors in parentheses; * p<0.10, ** p<0.05, *** p<0.010 S Notes: This table presents estimates of OLS regressions (column 1-2) and FE (column 3-4) regressions of the natural logarithm of night time lights at time t (a season in a year) on the operational status of the local market Market: Normal and Market : below with the reference group defined as closed during the period between 2015 and 2020. The controls included in the regression include: greenness as measured by the mean Normalized Difference Vegetation Index Mean ndvi, precipitation Mean precipitation and its square Mean precipitation2 and country × year dummies. Constants are not shown. 282 7.4 Results Technical Paper 6. Building Rural Development in the Lake Chad Region market status across the OLS (Columns 1–2) and FE specifications (Columns 3–4). The normal market status has a higher coefficient than below normal market status, which is inline with expectations. These results show that a normal operation status is associated with a 1.8 percent higher night time lights compared to the reference closed status. The natural logarithm to the nearest conflict event in the same season of the previous year is positive, whereby market locations farther from the conflict have on average higher night time lights. I find similar results for a one year lagged market status (results not shown). A report by Mercy Corps et al. (2017) stated that destroyed market outlets can typically take 9 to 12 months to reopen. Furthermore, findings from the report include 80 percent of interviewed farmers responded that their preferred or most frequented market closed during the insurgency where women and IDP and returnee farmers experienced a slightly higher incidence of market closures compared to the overall average. At the time of writing, the number of conflict events nearby Lake Chad continues at high levels and limits economic opportunities of displaced and conflict- affected households to earn income and now typically include firewood sales, petty trade, and construction labor (FEWS NET, 2021). 7.4 Results 283 Lake Chad Regional Economic Memorandum  |  Development for Peace 7.5 Conclusion Agriculture is important for the economies of the countries in the Lake Chad region. Farming, herding and fishing provide essential economic activity for many households. Using over three decades of remotely sensed and geospatial panel data to gain insight on agricultural activities, these results provide evidence that an increase in market access is associated with an increase in cultivated land and local agricultural economic activity. Given the modest increase in length of paved road during this period, I find the increase in market access is mainly driven by the growth in population rather than an improvement in roads, which is suggestive that growth in cultivated land is responsive to local demand similar to the findings in Berg et al. (2018). The findings of heterogeneous effects that an increase in market access in areas of shrinking cropland will further reduce cropland area is inline with Berg et al. (2018). Similarly, I find a positive association of market access with agricultural GDP using newly available local estimates. Although more land is under cultivation, the satellite derived measures of rainfed and irrigated cropland show little gains in irrigated land since 1992. It remains an important question to investigate for further examination whether or not gains in yield have corresponded commensurately with the increase in cultivated land. Even so, conflict can attenuate gains with negative impacts whereby the proximity of conflict events in the previous year lessens cropland expansion over the entire region and lessens night time lights in local markets. I find that the normal and below normal operational status of markets is associated with higher night time lights compared to closed markets. 284 7.5 Conclusion Technical Paper 6. Building Rural Development in the Lake Chad Region References Adebisi, SA, OO Azeez, and R Oyedeji, “Appraising the Effect of Boko Haram Insurgency on the Agricultural Sector of Nigerian Business Environment,” Journal of Law and Governance, 2016, 11 (1). Adelaja, Adesoji and Justin George, “Effects of conflict on agriculture: Evidence from the Boko Haram insurgency,”World Development, 2019, 117, 184–195. Aggarwal, Shilpa, “Do rural roads create pathways out of poverty? Evidence from India,” Journal of Development Economics, 2018, 133, 375 – 395. Agwu, George, “The Boko Haram conflict and food insecurity: Does resilience capacity matter?,” 2020. Ali, Rubaba, Alvaro Federico Barra, Claudia N Berg, Richard Damania, John Nash, and Jason Russ, Transport infrastructure and welfare: an application to Nigeria, The World Bank, 2015. Awodola, Bosede and Agyeno Oboshi, “Terrorism in Northern Nigeria: A threat to food security in Maiduguri,”Mediterranean Journal of Social Sciences, 2015, 6 (3 S2), 11–11. Bagozzi, Benjamin E, Ore Koren, and Bumba Mukherjee, “Droughts, land appropriation, and rebel violence in the developing world,” The Journal of Politics, 2017, 79 (3), 1057–1072. Barra, Federico, Claudia Berg, and Philip Verwimp, “Violent Conflict, Transport Costs, and Poverty: An instrumental variables approach with geospatial data for Nigeria,” mimeo. Beguería, Santiago, Sergio M Vicente-Serrano, Fergus Reig, and Borja Latorre, “Standardized precipitation evapotranspiration index (SPEI) revisited: parameter fitting, evapotranspiration models, tools, datasets and drought monitoring,” International journal of climatology, 2014, 34 (10), 3001–3023. Berg, Claudia N, Brian Blankespoor, and Harris Selod, “Roads and Rural Development in Sub-Saharan Africa,” The Journal of Development Studies, 2018, 54 (5), 856–874. __ , Uwe Deichmann, Yishen Liu, and Harris Selod, “Transport Policies and Development,” The Journal of Development Studies, 2017, 53 (4), 465–480. Blankespoor, Brian, Amjad Khan, and Harris Selod, “Consolidating Data of Global Urban Populations: a Comparative Approach,” AGUFM, 2017, 2017, IN52B–03. __ , Jedwab, Rémi, Brian Blankespoor, Takaaki Masaki, and Carlos Rodríguez-Castelán, “Estimating the Spillover Effects of Foreign Conflict: Evidence from Boko Haram,” forthcoming. __ , Sering Touray, and Roy Katayama, “Estimating the effect of conflict on agricultural activity in the Central African Republic with remotely sensed data,” forthcoming. __ , Yating Ru, Ulrike Wood-Sichra, Timothy S. Thomas, Liangzhi You, and Erwin Kalvelagen, “Gridded Agricultural GDP of theWorld,” forthcoming. Bruederle, Anna and Roland Hodler, “Nighttime lights as a proxy for human development at the local level,” PloS one, 2018, 13 (9), e0202231. Bucini, Gabriela and Eric F Lambin, “Fire impacts on vegetation in Central Africa: a remote-sensing-based statistical analysis,” Applied Geography, 2002, 22 (1), 27–48. Chamberlin, Jordan, TS Jayne, and Derek Headey, “Scarcity amidst abundance? Reassessing the potential for cropland expansion in Africa,” Food Policy, 2014, 48, 51–65. Crawford, Eric, Valerie Kelly, Thomas S Jayne, and Julie Howard, “Input use and market development in Sub-Saharan Africa: an overview,” Food Policy, 2003, 28 (4), 277–292. References 285 Lake Chad Regional Economic Memorandum  |  Development for Peace Damania, Richard, Claudia Berg, Jason Russ, A Federico Barra, John Nash, and Rubaba Ali, “Agricultural technology choice and transport,” American Journal of Agricultural Economics, 2017, 99 (1), 265–284. Dardel, Cécile, Laurent Kergoat, Pierre Hiernaux, E Mougin, Manuela Grippa, and CJ Tucker, “Re-greening Sahel: 30 years of remote sensing data and field observations (Mali, Niger),” Remote Sensing of Environment, 2014, 140, 350–364. de Haan, Leo, Paul Quarles van Ufford et al., “Cross-border cattle marketing in Sub-Saharan Africa since 1990: geographical patterns and government induced change,” 1999. Déby Itno, I., L. Fabius, Jean-Paul Moatti, N. Pourtier, and Christian Seignobos, Atlas of lake Tchad, Vol. number spécial 183 2015. Donaldson, Dave, “Railroads of the Raj: Estimating the impact of transportation infrastructure,” American Economic Review, 2018, 108 (4-5), 899–934. Eriegha, O, S Ovie, and A Ovie SOand Aminu, “Shrinking Lake Chad: Initialization of culture-based fisheries for improved livelihood in Nigeria,” International Journal of Fisheries and Aquatic Studies, 2019, 7 (3), 14–18. European Space Agency, “Land Cover CCI Product User Guide Version v2.0.7 Tech. Rep.,” 2017. __ , “C3S Global Land Cover products 2016-2018,” 2019. FAO, “Climate change predictions in Sub-Saharan Africa: impacts and adaptations (ClimAfrica),” 2013. __ , “Lake Chad Basin crisis - Response strategy (2017 - 2019),” 2017. FAO, “GIEWS Country Brief, Cameroon,” 2020. __ , “GIEWS Country Brief, Chad,” 2020. __ , “GIEWS Country Brief, Niger,” 2020. __ , “GIEWS Country Brief, Nigeria,” 2020. Federal Government of Nigeria, “Draft National Transport Policy,” Technical Report 2010. FEWS NET, “Production and Market Flow Map : Nigeria,” 2008. __ , “Production and Market Flow Map : West Africa,” 2009. __ , “Chad Rapid Livelihood Zone Profiles,” 2011. __ , “Niger Zoning Plus Activity Report,” 2011. __ , “Nigeria Food Security Outlook,” 2015. __ , “Nigeria Livelihood ZoneMap and Descriptions,” 2018. __ , “Cameroon Livelihood ZoneMap and Descriptions,” 2019. __ , “Nigeria Food Security Outlook,” 2019. __ , “Nigeria Food Security Outlook,” 2020. __ , “Nigeria Food Security Outlook,” 2021. Food and Agriculture Organization of the United Nations (FAO), “Crop Prospects and Food Situation,” 2020. Funk, Chris, Pete Peterson, Martin Landsfeld, Diego Pedreros, James Verdin, Shraddhanand Shukla, Gregory Husak, James Rowland, Laura Harrison, Andrew Hoell et al., “The climate hazards infrared precipitation with stations—a new environmental record for monitoring extremes,” Scientific data, 2015, 2 (1), 1–21. Gilbert, Marius, Gaëlle Nicolas, Giusepina Cinardi, Thomas P Van Boeckel, Sophie O Vanwambeke, GR William Wint, and Timothy P Robinson, “Global distribution data for cattle, buffaloes, horses, sheep, goats, pigs, chickens and ducks in 2010,” Scientific data, 2018, 5 (1), 1–11. Goldewijk, Kees Klein, Arthur Beusen, Jonathan Doelman, and Elke Stehfest, “New anthropogenic land use estimates for the Holocene: HYDE 3.2,” Earth System Science Data, 2017, 9 (2), 927–953. 286 References Technical Paper 6. Building Rural Development in the Lake Chad Region Guenang, Guy Merlin and F Mkankam Kamga, “Computation of the standardized precipitation index (SPI) and its use to assess drought occurrences in Cameroon over recent decades,” Journal of Applied Meteorology and Climatology, 2014, 53 (10), 2310–2324. Harris, Ian, Timothy J Osborn, Phil Jones, and David Lister, “Version 4 of the CRU TS monthly high-resolution gridded multivariate climate dataset,” Scientific data, 2020, 7 (1), 1–18. Henderson, J Vernon, Tim Squires, Adam Storeygard, and David Weil, “The Global Distribution of Economic Activity: Nature, History, and the Role of Trade,” The Quarterly Journal of Economics, 09 2017, 133 (1), 357–406. Henderson, Vernon, Adam Storeygard, and David N Weil, “A bright idea for measuring economic growth,” American Economic Review, 2011, 101 (3), 194–99. Jedwab, Rémi and Adam Storeygard, “Economic and political factors in infrastructure investment: evidence from railroads and roads in Africa 1960–2015,” Economic History of Developing Regions, 2019, 34 (2), 156–208. __ and __ , “The Average and Heterogeneous Effects of Transportation Investments: Evidence from Sub-Saharan Africa 1960-2010,”Working Paper 27670, National Bureau of Economic Research August 2020. Jelilov, Gylych, Ramat Ayinde, Selman Tetik, Bilal Celik, and Natalia Olali, “Impact of Terrorism on Agricultural Business in Borno State, Nigeria,” Applied Studies in Agribusiness and Commerce, 2018, 12 (1033-2019- 3300), 117–124. Kull, Christian A and Paul Laris, “Fire ecology and fire politics in Mali and Madagascar,” in “Tropical fire ecology,” Springer, 2009, pp. 171–226. Lehner, Bernhard and Petra Döll, “Development and validation of a global database of lakes, reservoirs and wetlands,” Journal of hydrology, 2004, 296 (1-4), 1–22. Lemoalle, Jacques and Sanusi Imran Abdullahi, eds, Le développement du lac Tchad/Development of Lake Chad: Situation actuelle et futurs possibles/Current Situation and Possible Outcomes, IRD Éditions, 2017. __ , Jean-Claude Bader, Marc Leblanc, and Ahmed Sedick, “Recent changes in Lake Chad: Observations, simulations and management options (1973–2011),” Global and Planetary Change, 2012, 80, 247–254. Li, Xuecao, Yuyu Zhou, Min Zhao, and Xia Zhao, “A harmonized global nighttime light dataset 1992–2018,” Scientific Data, 2020, 7 (1), 1–9. Liu, Xiaoxuan, Le Yu, Wei Li, Dailiang Peng, Liheng Zhong, Le Li, Qinchuan Xin, Hui Lu, Chaoqing Yu, and Peng Gong, “Comparison of country-level cropland areas between ESA-CCI land cover maps and FAOSTAT data,” International Journal of Remote Sensing, 2018, 39 (20), 6631–6645. Magrin, Géraud, Marc-Antoine Pérouse de Montclos, Christian Seignobos, and Pauline Gluski, “Crise et développement: la région du lac Tchad à l’épreuve de Boko Haram,” 2018. Maystadt, Jean-François, Valerie Mueller, Jamon Van Den Hoek, and Stijn Van Weezel, “Vegetation changes attributable to refugees in Africa coincide with agricultural deforestation,” Environmental Research Letters, 2020, 15 (4), 044008. Mercy Corps, USAID, Oxfam, Action against Hunger, CRS, Rescue, and COPI, “Northeast Nigeria joint livelihood and market recovery assessment,” 2017. Motta, Paolo, Thibaud Porphyre, Ian G Handel, Saidou M Hamman, Victor Ngu Ngwa, Vincent N Tanya, Kenton L Morgan, and B Bronsvoort, “Characterizing Livestock Markets, Primary Diseases, and Key Management Practices Along the Livestock Supply Chain in Cameroon,” Frontiers in veterinary science, 2019, 6, 101. __ , __ , Ian Handel, Saidou M Hamman, Victor Ngu Ngwa, Vincent Tanya, Kenton Morgan, Rob Christley, M Barend et al., “Implications of the cattle trade network in Cameroon for regional disease prevention and control,” Scientific reports, 2017, 7, 43932. References 287 Lake Chad Regional Economic Memorandum  |  Development for Peace Nachtergaele, Freddy, Riccardo Biancalani, Sally Bunning, Hubert George et al., “Land degradation assessment: the LADA approach,” in “in” 2010. NASA, “MCD64A1.006 MODIS Burned Area Monthly Global,” 2020. __ , “MOD13A2.006 Terra Vegetation Indices 16-Day Global 1km ,” 2020. __ , “MOD17A2H.006: Terra Gross Primary Productivity 8-Day Global,” 2020. National Geospatial-Intelligence Agency, “World Port Index,” 2014. Nilsson, Erik, Stefan Hochrainer-Stigler, Junko Mochizuki, and Cintia Bertacchi Uvo, “Hydro-climatic variability and agricultural production on the shores of Lake Chad,” Environmental Development, 2016, 20, 15–30. OCHA, “Lake Chad Basin Humanitarian Snapshot - September 16,” 2020. Odada, E, L Oyebande, and J Oguntola, “Experiences and lessons learned: Brief for Lake Chad,” Global Environment Facility (Lake Basin Management Initiative)-International Waters Learning Exchange and Resource Network (IW: LEARN), 2003. Odada, Eric O, Lekan Oyebande, and AJ Oguntola, “Lake Chad: experience and lessons learned brief,” 2005. OECD, “Atlas on Regional Integration in West Africa,” 2006. [Map 8; accessed September 20, 2020]. Okpara, Uche T, Lindsay C Stringer, Andrew J Dougill, and Mohammed D Bila, “Conflicts about water in Lake Chad: Are environmental, vulnerability and security issues linked?,” Progress in Development Studies, 2015, 15 (4), 308–325. Raji, A, “The past history and present trends in the fisheries of Lake Chad,” 1993. Raleigh, Clionadh, Andrew Linke, Håvard Hegre, and Joakim Karlsen, “Introducing ACLED: an armed conflict location and event dataset: special data feature,” Journal of peace research, 2010, 47 (5), 651–660. Redding, Stephen J and Matthew A Turner, “Transportation costs and the spatial organization of economic activity,” in “Handbook of regional and urban economics,” Vol. 5, Elsevier, 2015, pp. 1339–1398. Running, Steven W, Ramakrishna R Nemani, Faith Ann Heinsch, Maosheng Zhao, Matt Reeves, and Hirofumi Hashimoto, “A continuous satellite-derived measure of global terrestrial primary production,” Bioscience, 2004, 54 (6), 547–560. Salemi, Colette, “Refugee camps and deforestation in Sub-Saharan Africa,” Journal of Development Economics, 2021, p. 102682. Sarch, M-T, “Fishing and farming at Lake Chad: Institutions for access to natural resources,” Journal of Environmental management, 2001, 62 (2), 185–199. Stauch, A., “Développement et évolution de la pêche dans le bassin tchadien,” 1960. Storeygard, Adam, “Farther on down the Road: Transport Costs, Trade and Urban Growth in Sub-Saharan Africa,” The Review of Economic Studies, 04 2016, 83 (3), 1263–1295. Strobl, Eric and Robert O Strobl, “The distributional impact of large dams: Evidence from cropland productivity in Africa,” Journal of development Economics, 2011, 96 (2), 432–450. Thomas, Timothy, Liangzhi You, Ulrike Wood-Sichra, Yating Ru, Brian Blankespoor, and Erwin Kalvelagen, “Generating Gridded Agricultural Gross Domestic Product for Brazil: A Comparison of Methodologies,” Technical Report, The World Bank 2019. Tobler,Waldo, “Three presentations on geographical analysis and modeling,” 1993. UNHCR and World Bank, “Forced Displacement by the Boko Haram Conflict in the Lake Chad Region,” 2016. USAID, “Lake Chad Basin - Complex Emergency,” 2020. Van Den Hoek, Jamon, “Agricultural market activity and Boko Haram attacks in northeastern Nigeria,” OECD Working Paper 9 2017. 288 References Technical Paper 6. Building Rural Development in the Lake Chad Region Verdin, Kristine L, Jonathan W Godt, Chris Funk,Diego Pedreros, Bruce Worstell, and James Verdin, “Development of a global slope dataset for estimation of landslide occurrence resulting from earthquakes,” Technical Report 2007. Weide, Roy Van Der, Bob Rijkers, Brian Blankespoor, and Alexei Abrahams, “Obstacles on the road to Palestinian economic growth,” 2018. Weiss, Marie, Frédéric Jacob, and G Duveiller, “Remote sensing for agricultural applications: A meta-review,” Remote Sensing of Environment, 2020, 236, 111402. WFP, “Regional Market Assessment,” 2016. Yu, Qiangyi, Liangzhi You, Ulrike Wood-Sichra, Yating Ru, Alison KB Joglekar, Steffen Fritz, Wei Xiong, Miao Lu, Wenbin Wu, and Peng Yang, “A cultivated planet in 2010: 2. the global gridded agricultural production maps,” 2020. Zaveri, Esha, Jason Russ, and Richard Damania, “Drenched fields and parched farms: Evidence along the extensive and intensive margins,” Available at SSRN 3104224, 2018. __ , __ , and __ , “Rainfall anomalies are a significant driver of cropland expansion,” Proceedings of the National Academy of Sciences, 2020, 117 (19), 10225–10233. References 289