Climate shocks, cash crops and resilience: Evidence from colonial tropical Africa.  Kostadis J. Papaioannou Wageningen & Utrecht University [email protected] & Michiel de Haas Wageningen University [email protected] October 2015 Abstract. A rapidly growing body of research examines how weather variability, anomalies and shocks influence economic and societal outcomes. This study investigates the effects of weather shocks on African smallholder farmers in British colonial Africa and intervenes in the debate on the mediating effect of cash crops on resilience to shocks. We employ a dual research strategy, involving both qualitative and econometric analysis. We analyse original primary evidence retrieved from annual administrative records and construct a panel dataset of 151 districts across West, South-central and East Africa in the Interwar Era (1920-1939). Our findings are twofold. First, we qualitatively expose a range of mechanisms leading from drought and excessive rainfall to harvest failure, social tension and distress. We then test the link econometrically and find a robust and significant U-shaped relation between rainfall deviation and social tension and distress, proxied by annual imprisonment. Second, we review a long-standing and unsettled debate on the impact of cash crop cultivation on smallholders’ resilience to climatic shocks and find that cash crop districts experienced lower levels of social tension and distress in years of extreme rainfall variability.  Acknowledgments: We are grateful to Peter Bent, Bram van Besouw, Erwin Bulte, Cédric Chambru, Daniel Curtis, James Fenske, Ewout Frankema, Jose Martinez-Gonzalez, Leander Heldring, Paolo Malanima, Pieter Woltjer, and the participants of the International Workshop in Historical Ecology at Uppsala University (November 2014), the Thesis Workshop Series in London School of Economics and Political Science (LSE) (February 2015), the seminar of Political Economy, Economic History, Growth and Development at Groningen University (February 2015), the Economic History Graduate seminar in Oxford (May 2015), Datini-ESTER advanced seminar in Prato, Italy (May 2015) and XIIth World Economic History Congress in Kyoto, Japan (August 2015). We are grateful for the financial support of the European Research Council under the European Community’s Seventh Framework Programme (ERC Grant Agreement no. 313114) as part of the project Is Poverty Destiny? A New Empirical Foundation for Long-Term African Welfare. The usual disclaimer applies. 1 1. Introduction “...the year had gone mad. Rain fell as it had never fallen before. For days and nights together it poured down in violent torrents, and washed away the crops [...] The blazing sun returned, more fierce than it had ever been known, and scorched all the green that had appeared [...] He watched the sky all day for sign of rain clouds and lay awake all night. In the morning he went back to his farm and saw the withering tendrils.” Chinua Achebe [When Things Fell Apart (1994) p. 16] Over the past years, scholars and policy makers have become increasingly aware of the short- and long-run impact of climatic factors on economic, social and political outcomes (Hsiang et al. 2013; Dell et al. 2014). The impact of rainfall on society is particularly pronounced in developing countries, with sub-Saharan Africa being the most vulnerable region (Ahmed et al. 2009; Barrios et al. 2010; Bruckner & Ciccone 2011; Miguel et al. 2004). This is not surprising, as a large share of the population depends on rain-fed subsistence agriculture and less than 5% of the cultivated area is irrigated (Schlenker & Lobell 2010). Indeed, in an African context, harvest failure and food insecurity arising from climatic factors appear to be tightly related to social destabilization. Several studies have shown that climatic factors trigger social conflict (Fjelde & von Uexkull 2012; Hendrix & Salehyan 2012) and full-blown civil war (Blattman and Miguel, 2010), but also processes of democratization (Bruckner & Ciccone 2011). Rather than accepting a static link between climate and societal outcomes, a key challenge is to understand which conditions aggravate or mediate the impact of shocks. It is especially crucial to learn more about the local determinants of resilience to the vagaries of climate (Adger 2000; Folke 2006; Gallopin 2006). Africa’s rising population densities, enveloping climate change and resurging socio-political instability contribute to making this a most pressing concern. This study offers a number of contributions. First, it provides novel evidence on the impact of weather shocks on societal outcomes in tropical Africa from a historical perspective. While this link is subject to a wide range of studies, the number of sources is relatively thin, and the debate far from settled (Dell et al. 2014; Hsiang et al. 2013; Klomp & Bulte 2013). Exploiting the extensive and consistent administrative records that remain from Britain’s African empire, we provide new material on a region for which systematic data collection is notoriously difficult. Moreover, our focus on the interwar era (1920-1939) contributes to a considerable expansion of the time horizon (cf. Papaioannou 2014; Christian & Fenske 2015). Our argument is based on both qualitative evidence and econometric analysis. Initially, we use colonial administrative accounts to expose the mechanisms that lead from extreme weather shocks to higher levels of social distress. We find that, 2 in line with the environmental scarcity literature, these accounts strongly suggest that extreme weather events bring about crop and harvest failures which, in turn, increase competition over ever scarce resources and ultimately lead to higher levels of social tension (Homer-Dixon 1999, Kahl 2006). Interestingly, they also suggest that both drought and excessive rainfall adversely affect agricultural outcomes. To test the rainfall-to-distress link econometrically, we construct a novel panel dataset at the district-level obtaining observations on annual rainfall and imprisonment –our proxy for social tension and distress; our dataset consists of 151 districts for a 20 year period. Running a dynamic panel data specification (using system-GMM), with both time and district fixed effects, we find a robust u-shaped effect of weather shocks on social tension and distress. Second, this study investigates the extent to which resilience to weather shocks is mediated by smallholders’ cash crop cultivation. The adoption of cash crops has been variously portrayed as either beneficial or detrimental for African livelihoods (Hopkins 1973; Myint 1958; Rodney 1978; Maxwell & Fernando 1989; Austin 2014). The issue whether the introduction of cash crops was beneficial, or detrimental to vulnerable rural communities is the subject of a multifaceted, heated and long-standing debate among policy makers and social scientists. These opposed views also apply to the impact of cash crops on rural communities’ ability to cope with shocks. Some maintain that the introduction of cash crops made rural communities more vulnerable, by diverting attention away from subsistence production, undermining ‘traditional insurance mechanisms’ and facilitating exploitation and extraction (Vaughan 1987; Watts 1983). Others, instead, have argued that access to export markets made such communities more resilient, providing them with the ability to spread risk, smoothen consumption and profit from infrastructural and institutional investments (Bryceson, 1980, 1988; Fafchamps 1992b; Maxwell & Fernando 1989). We argue that districts with lower spikes of social distress in years of weather shocks are more resilient, i.e. have stronger abilities to deal with shortages and absorb shocks (Adger 2000; Folke 2006; Gallopin 2006). Based on primary sources, we propose a range of mechanisms that reveal the mitigating effect of cash crops on resilience. Our econometric analysis corroborates this argument. To the best of our knowledge, we are the first to construct a new cross-sectional, district- level index which captures smallholders’ involvement in the export crop economy; we named this index cash crop intensity. We demonstrate that districts with relatively higher levels of cash crop cultivation were significantly more resilient to weather shocks than those with less cash crops. We 3 perform a number of robustness tests to explore the extent to which the link between cash crop cultivation and resilience is causal (‘horse-race’ tests). We justify our case study of interwar British colonial Africa on a number of grounds.i First, some parts of British colonial Africa had already experienced considerable smallholder-based agricultural commercialization by this period, whereas others did not. Besides some areas which were dominated by settlers and an emerging mining sector, smallholders’ cash crops dominated export trade. In this way, we are able to exploit the variation necessary to test the heterogeneous impact of rainfall shocks. Second, the interwar period was the most tranquil and peaceful one within the era of colonial rule (Killingray 1986). Thus, linking levels of social tension and distress to harvest failures (rather than political upheaval) is more possible than during the period of violent early-colonial conquest and the highly politicized post-war road to independence. Third, Britain administered a vast African empire. The colonial state’s key preoccupations related to law and order as well as agricultural production, coinciding with our key variables of interest. The extensive bureaucratic legacy has allowed us to construct a consistent district-level dataset spanning approximately one-fifth of Africa’s landmass and one-third of its population in this period. The remainder of the paper is organized as follows. Section two deals with the impact of climate shocks on social tension and distress. It introduces the literature and provides evidence, based on original qualitative sources, to reason that rainfall shocks are tightly related to harvest failures, and in turn to social tension and distress; it justifies the key variables, it formulates the testable hypotheses and it presents the main results on the robust curvilinear (U-shaped) relation between weather shocks and social tension. Section three follows a similar structure. It provides both qualitative and quantitative evidence to show that districts with cash crops were more resilient to weather-induced scarcities, and argues that this effect is likely to be driven to a considerable extent by cash crops. Section four concludes and suggests directions for further research. 2. Do weather shocks lead to social tension and distress? 2.1 The debate Over the past decade, the scholarly debate on the societal impact of climate on societal outcomes has expanded considerably. Weather variables have been linked to economic outcomes, health, agriculture, crime and conflict (for a good overview of this body of literature, see Hsiang et al. 2013; Dell et al. 2014). Whereas such an effect of weather may not always be present in wealthy, stable countries (Dell, Jones & Olken 2012), it seems particularly pronounced in tropical Africa 4 (Barrios et al. 2010; Bruckner & Ciccone 2011; Miguel et al. 2004; Papaioannou 2015). While some scholars dispute the evidence linking climate to conflict (Klomp & Bulte, 2013), most find support for the existence of a causal relation, especially in low-income settings (Hsiang et al. 2013). Moreover, the literature has been radically divided with regard to the mechanisms that explain the climate-to-conflict nexus (Buhaug 2010; Klomp & Bulte 2013). To resolve such issues, the literature would profit from more fine-grained analysis and an extension of the time period. Among those who take precipitation as the key independent variable, some find that drought is the prime driver of conflict (Maystadt & Ecker 2014; Christian & Fenske 2015), while others argue that the relationship is u-shaped, with extremes on both ends (drought and excessive rainfall) leading to higher conflict incidence (Papaioannou 2014; Fjelde & von Uexkull 2012; Hendrix & Salehyan 2012). The impact of weather deviations on societal outcomes runs through an intermediate mechanism, for which harvest failure (leading to lower incomes and agricultural deficiencies) appears to be a prime candidate, especially in low-income settings, where people’s livelihoods tend to rely more heavily on (rain-fed) farming and where small deviations in crop yields can have devastating effects (Barrios et al. 2010; Bruckner & Ciccone 2011; Miguel et al. 2004; Schlenker & Lobell 2010). However, the debate on mechanisms to explain the weather- conflict link is not resolved. The impact of extreme climatic anomalies on economic outcomes can be framed in terms of an opportunity cost model. One can argue that, as (agricultural) productivity declines as a result of climatic anomalies, engaging in unrest or crime becomes more opportune relative to participating in ‘peaceful’ economic activities. At the same time, however, it has been argued that abundance, rather than scarcity, encourages conflict, as there is more to gain from conflict in abundant rather than meagre years (Witsenburg & Adano 2009). Moreover, some argue that certain weather conditions affect behaviour directly, through psychological mechanisms (see for references Crost et al. 2015; Anderson 1989). The impact of climate can also be explained in terms of an environmental scarcity perspective, which sets out to explain how weather anomalies disturb people’s livelihoods, creating conditions that are more prone to increase social tension. Such conditions include population movements and intensified competition over scarce resources, such as wells and pasture land (Homer-Dixon, 1999; Kahl, 2006). In an African context, numerous studies have found that weather extremes lead to tension and conflict (Almer & Boes 2012; Buhaug 2010; Burke et al. 2009; Couttenier et al 2011; Fjelde & von Uexkull 2012; Hendrix & Salehyan 2012; Miguel et al. 2004). Most studies use a binary 5 indicator of conflict, for example based on the ACLED database which records conflict events in states affected with civil war (1960 – 2004), or the UCDP/PRIO Armed Conflict Dataset on armed conflict (1946 – present), which captures only large-scale outbreaks of unrest, violence or conflict, and omits subtler forms of social tension and distress. A number of recent contributions have begun to investigate different time periods (Papaioannou 2015, Christian & Fenske 2015), employ a more fine-grained, sub-national scope (Harari et al. 2013; Papaioannou 2015; Raleigh & Urdal 2007; cf. Jia 2014) and use more non-binary indicators of conflict intensity (Papaioannou 2015). Moreover, a number of studies have employed detailed case study analyses to uncover the key mechanisms leading from weather to conflict (Adano et al. 2012, Benjaminsen et al. 2012; Ember et al. 2012; Witsenburg & Adano 2009). We contribute to these new explorations by providing a new indicator on more subtle forms of social tension and distress. 2.2 Qualitative evidence: 2.2.a Rainfall shocks and harvest failures under British colonialism Our sources provide a unique opportunity to engage with the different perspectives in the debate on the effects of weather anomalies on African societies. The British colonizers set up an extensive system of administration in their African dependencies. Territories were subdivided into provinces and districts. Local administration was left to African native rulers, who operated under supervision of British administrative officers.ii Elaborate administrative accounts were kept. Local officers reported on a regular basis to their superiors on a range of issues. Each colony produced a series of departmental annual reports concerning a wide range of issues. In this paper, we use such annual reports obtained from the departments of agriculture, native affairs, police, justice and prisons, as well as the annual Blue Books of Statistics. These reports are rather consistent in their coverage of issues over time and across colonies and give us a uniquely comprehensive insight into local conditions across a wide area and a considerable timespan. The goal of this section is to i) investigate whether the relationship between weather and agricultural outcomes should be conceptualized as linear or u-shaped, and ii) explore the mechanisms that explain the relationship. The impact of weather conditions on agricultural outcomes is extensively discussed by colonial administrators. Reports make regular notice of weather induced agricultural failure, resulting in higher levels of distress, and in the more extreme cases, subsistence crises and famines. Frequently mentioned negative results of droughts include: 6 (a) Crop damage and failure: In a context of rain-fed agriculture, lack of precipitation prevents seeds from germinating, slows down plant growth and diminishes yields.iii In severe cases, drought is also associated with complete crop failure, dust storms and soil erosion. (b) Livestock starvation: Drought diminishes water supply to wells, leads surface water and pastures to dry up, negatively impacting upon the water and food supply of livestock.iv To prevent starvation, cattle has to be moved, which in turn increased their susceptibility to disease and further weakens underfed herds. Regular mention is also made of the adverse effects of excessive rainfall on agriculture. The adverse effects of excessive rainfall run via a number of specified mechanisms: (c) Crop damage and failure: Excessive precipitation and subsequent flooding has the potential to seriously damage, or even completely destroy the harvest.v (d) Worsened phytopathological conditions: Heavy rainfall and the resulting humidity increases the likelihood of outbreaks of disease,vi the spread of parasitic organisms,vii and the arrival of locusts (the latter especially when rainfall has been preceded by a drought spell). (e) Logistical problems: Excessive rainfall impedes farmers’ from cultivatingviii, storingix and transportingx their produce. It is important to note, however, that abundant rainfall, albeit in a much more limited number of cases, also seems to have had benevolent effects. In some cases, above average rainfall appears to have been ‘generous’ rather than ‘excessive’, bringing about good harvests and abundant pasturexi, and feeding rivers that could be used for irrigation and transport.xii Interestingly, excessive rainfall at times resulted in diametrically opposed effects simultaneously. Plenty rainfall could lead to bountiful harvests and plentiful pasture in one district, while causing floods and crop failure in a neighbouring district.xiii Although the effect of excessive rainfall was not entirely unambiguous, mechanisms (c), (d) and (e) were much more commonly noted, suggesting that agricultural outcomes were not only adversely affected by droughts but also by excessive rain. It is important to note in this context that British colonial Africa encompassed a wide variety of agro-ecological settings, including arid regions, as well as areas with abundant rainfall.xiv Examples of the negative impact of droughts as well as excessive rainfall can be found in both very wet and very dry regions. Even though dry districts, in some cases, seem to have profited from a season of plenty rainfall leading to an extension of pastures and cropland, heavy showers are also reported to have had negative results, for example in cases dry soils were incapable of absorbing the precipitation, resulting in run-off, floods and waterlogging.xv Hence, we conclude that our 7 qualitative evidence supports a U-shaped conceptualization of the link between weather and agricultural outcomes.1 2.2.b From harvest failures to social tension and distress Once we have established that rainfall deviations led to diminished harvest or even provoked failure, we should investigate to what extent these agricultural outcomes, in their turn, provoked social tension and distress. Turning to the source material, we observe and propose several plausible mechanisms to argue that the recorded impact of extreme rainfall deviations on agricultural production was indeed substantial for local communities and resulted in actual hardship: (f) Scarcity, price spikes and speculation: Harvest failures are frequently reported to have resulted in spectacular food prices hikes and resultant social tension and distress.xvi Likewise, colonial officers sometimes explicitly attributed the absence of tension to low prices due to favourable weather.xvii It is important to note that such price spikes were not always the result only of environmental scarcities, but were aggravated by human intervention. In some instances of (impending) harvest failures, speculators were said to stock food crops to further drive up prices and make high profits.xviii Speculation had the potential to cause distress and to certainly increase levels of social tension.xix Drought-induced stock mortality resulted in an increased incidence of petty crime, as well as more serious stock raids and thefts in adjacent areas.xx In extreme cases, rainfall-induced scarcity was so serious as to result in fatal local famines.xxi (g) Population movements: Adverse weather conditions forced people to get on the move. In some cases such movements were motivated by a desire to earn an income elsewhere to compensate for lost harvestsxxii or to find pastures for their cattle.xxiii In other, more acute cases people wandered in search of food, out of pure desperation or because severe flooding had destroyed their homes.xxiv Pastoralist groups were reported to engage in drought-induced migration, which carried an increased risk of clashes with neighbouring groups over grazing pastures and water.xxv In some cases, violent clashes indeed seem to have erupted,xxvi while in other cases they were reported to have been averted.xxvii (h) Unrest: Conflicts arising out of dissatisfaction of local groups with colonial authorities have also been argued to have erupted due to drought conditions.xxviii 1 It should be mentioned here that both positive and negative deviations from any long-term rainfall mean affect farmers across a wide range of agro-ecological settings is not surprising, especially when realizing that smallholders build their farming systems around an expected level of rainfall. The choice of the crop mix and farming methods are calibrated on the basis of this expectation. 8 These mechanisms indicate that weather fluctuations could have a direct impact upon human societies. At the same time, however, we must also acknowledge that we find some instances where the administrators link depressed incomes to lower, rather than higher, levels of social tension, as adverse conditions forced people to focus all their attention to agriculture.xxix 2.3 Imprisonment as a measure of social tension and distress While the administrative accounts are rich and enable us to identify a range of mechanisms running from rainfall deviations through agricultural outcomes to social tension and distress, they suffer from considerable limitations. British colonial services tended to be understaffed, and local administrators had to operate on a shoestring and were hardly capable to administer the vast territories they were supposed to control.xxx Moreover, their accounts reveal strong prejudicexxxi and paternalisticxxxii and derogatoryxxxiii attitudes towards local populations. On top of that, previous scholars have pointed out that civil servants, to brush up their achievements and benefit their own careers, had incentives to focus on ‘progress’ and paint a rosy picture to superiors, which may have affect the reliability of local administrative accounts (Killingray 1986).xxxiv As a result of these factors, coverage of events may be incomplete and their representation biased. For this reason, we have found it necessary to identify a more objective and consistent district-level indicator of social tension and distress. In an African (historical) context, such indicators are scarce. Annual crop yields, rural food prices, or statistics on mortality and fertility are not available and most of the times inconsistently reported. Data on conflict is scarce as well. Notable exceptions are the studies of Huillery (2011), who collected data on conflict, using colonial reports to sample binary district- level data on unrest in French West Africa for 16 years between 1906 and 1956 and that of Papaioannou (2014), who collected annual province-level indicators of conflict (imprisonment, court cases and homicides) in colonial Nigeria for 33 years between 1912 and 1945. We argue that imprisonment statistics, which are consistent and annually reported, could serve as a highly suitable proxy for social tension and distress. In the following paragraphs, we will argue how colonial-era imprisonment figures inform us about levels of social tension and distress. One of the key aspects of colonial rule consisted of the establishment of law and order. The colonial penal institutions, grafted upon penal codes and using imprisonment, fines and, in some cases corporeal punishment, as its major instruments, were essentially foreign to most parts of sub- Saharan Africa (Milner, 1969; Killingray, 1986; Bernault 2003, 2007). Once colonial penal systems were established, they served both to maintain social order and to strengthen colonial domination (Bernault 2007; Hynd 2011; Killingray 1986). 9 The newly established penal systems were used to incarcerate people for a range of reasons. Most imprisonments resulted from crimes related to theft or offences against the person.xxxv A considerable number of imprisonments were the result of debt and tax defaulting, and transgression of a range of colonial ordinances (Hynd 2011). Prison was not the only possible punishment for crime. Death penalties existed for the most serious crimes, and in some places corporeal punishments were common (Hynd 2008). Moreover, fines, being a much cheaper punishment for minor offences, were regularly awarded. In case a convicted person was unable to pay the fine, the sentence would be commuted into imprisonment (McCracken 1986; Coldham 2000). In most cases, administering justice (i.e. applying the penal code) happened at the discretion of the district officer (who had enjoyed only minor legal training) or, for minor crimes, native authorities (Milner 1969). As such, penal systems were sensitive to abuse. In settler dominated territories such as Southern Rhodesia and Kenya the penal systems partly served the interests of expatriate farmers to discipline labour and reinforce their property rights (Anderson 1993; Deflem 1994; Branch 2005). In other places, the majority of imprisonments resulted from trials in relatively autonomous native courts, which were, in their turn, not entirely free from manipulation by local elites (Killingray 1986; McCracken 1986). Moreover, some authors argue that colonial states used imprisonment to generate cheap convict labour (Bernault (2003; 2007). In general, the definition of ‘crime’ is highly contentious in a colonial setting with imposed ‘alien’ penal institutions (Kercher 1981). At least a share of the cases of imprisonment resulted from civil disobedience and rebellion rather than activities that were, at the time, commonly accepted by African populations as crimes (Branch 2005). Incarceration rates varied considerably across districts and colonies, reflecting differences in levels of coercion and state capacity. It is important to stress that our analysis does not intend to capture differences in absolute levels of imprisonment across colonial districts, but to use annual fluctuations of imprisonment to investigate the match between conflict spikes and rainfall shocks.2 Whereas it is doubtful if colonial imprisonment figures capture crime or conflict, we argue that annual imprisonment fluctuation is a highly suitable proxy for social tension and distress. We argue that all of the reasons for imprisonment identified above increase in years of distress, while additional reasons for imprisonment arose in such years: 2 Some scholars have pointed out that in the long run high absolute levels of social tension and unrest may provoke democratic transitions (Bruckner & Ciccone 2011). In an African context, conflict rose during the post-war road to independence such as the Mau Mau uprising in Kenya, and the Nyasaland emergency in 1959. However, such long-run outcomes of conflict are beyond the scope of this paper. 10