![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
A number of studies have examined the link between rainfall and conflict but results so far have been inconclusive. This study examines the effect of rainfall on economic performance in different sectors and conflict onset. The empirical analysis finds no support for a strong relation between rainfall and conflict as most results are not robust to different model specifications. The results also do not provide conclusive evidence for a link between growth in specific economic sectors and civil conflict onset.
Keywords:
Acknowledgements
Many thanks to Edwin de Jonge from Statistics Netherlands and David Bolvin from NASA-GPCP for their help with the rainfall data. I also would like to thank one anonymous referee for useful recommendations, and Mike Spagat from Royal Holloway for helpful discussions.
Notes
1 Due to the innovative approach their paper has quickly become one of the standard works in the literature. It has been cited 233 times since it was published in 2004 (number based on statistics from Web of Knowledge in July 2013).
2 The Journal of Peace Research published a special edition on climate change in January 2012 on the link between climate change and conflict containing three papers looking at rainfall all with different results.
3 The paper by Djankov and Reynal-Querol (Citation2010) is one of the few that contradicts the conclusion that poverty is a main determinant of conflict.
4 A number of papers have criticised the results of MSS on basis that their model is agnostic to the sector experiencing the shock (Dunning Citation2008), the fact that they erroneously code the dependent variable to include countries participating in civil wars in other states (Sandholt Jensen and Skrede Gleditsch Citation2009), and most forcefully by Ciccone (Citation2011) who argues that the results are driven by a positive correlation between conflict in t and rainfall in t – 2.
5 In this short review, I focus on the papers that have looked at the link between rainfall and conflict. For an insight into the relation between rainfall and economic growth I refer to (Barrios, Bertinelli, and Strobl Citation2010); Dell, Jones, and Olken (Citation2012).
6 Koubi et al. (Citation2012) looked at data for the whole world but found similar results for Africa alone.
7 Data can be found at http://www.pcr.uu.se/research/ucdp/datasets/ucdp_prio_armed_conflict_dataset/. Civil conflicts are all observations coded as type 3 or 4 in the data-set. A type 3 conflict is defined as ‘an internal armed conflict that occurs between the government of a state and one or more internal opposition group(s) without the intervention from other states’. A type 4 conflict is similar only allowing for the intervention from other states on behalf of one or both sides (Gleditsch et al. Citation2002).
8 A conflict is defined as: ‘a contested incompatibility that concerns government and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related deaths.’
9 The conflict incidence indicator is coded as:
10 An additional problem is the systematic measurement error in conflict countries due to the collapse of public institutions that provide the data.
11 The data-set is weakly unbalanced due to the fact that Namibia & Eritrea gained independence in 1990 and 1993, respectively. The model estimations include 46 countries due to the lack of data for Somalia.
12 I use GDP in constant US$ in order to capture international purchasing power.
13 Data in ASCII format is available at ftp://rsd.gsfc.nasa.gov/pub/912/bolvin/GPCP_ASCII/. For further information on the data itself, see project page: http://precip.gsfc.nasa.gov/.
14 Typically using GIS software, a value in a grid-cell is aggregated when the cell’s centre-point falls within the national boundaries.
15 Many thanks to Robert Hall from the South East Asia Research group at Royal Holloway for suggesting this method.
16 At the original resolution, aggregating returns no estimates for the following countries: Burundi, Cape Verde, Comoros, Djibouti, Gambia, Guinea-Bissau, Lesotho, Mauritius, Rwanda, Sao Tome & Principe. For the Seychelles, the higher resolution aggregation method does also not return an estimate.
17 Hendrix and Glaser (Citation2007) use a similar method, they refine the raster to 1 × 1 degree resolution and assign cells to countries on the basis of majority.
18 Note that each of the smaller grid-cells has the same value as the larger original grid-cell. This approach does not change anything about the measurements and therefore does not take into account any within-cell variation.
19 Rainct = dJanct × 31 + dFebct × 28 + … + dDecct × 31.
20 The time-series correlation is 0.99.
21 The high correlation between the two methods of aggregation can be explained by the fact that most African countries have very arbitrary boundaries. At the 1884–1885 Berlin conference, a majority of borders were drawn along meridians and parallels. Fourty-four percent of African borders follow meridians and parallels while 30% follow other rectilinear or curved lines (Alesina, Easterly, and Matuszeski Citation2008; Posner Citation2006).
22 Long term is in this case the period from 1979 to 2010.
23 Four hundred and ninety observations (35%) are both positive and 527 observations (38%) are both negative. In 188 observations, where anomalies have a positive sign, the growth rate is negative (14%) and for 184 observations, where the deviations from the panel mean are negative, the growth rate is positive (13%).
24 The specific set of controls that are used in this study are based on the papers by Miguel, Satyanath, and Sergenti (Citation2004) and Fearon and Laitin (Citation2003).
25 Peace years are thus measured as the absence of any violence that resulted in at least 25 battle-related deaths in a given year.
26 Or otherwise, the time since 1945 though this did not occur in the panel.
27 Anocracies are more vulnerable due to political instability, the result of the spread of power among elitist groups constantly competing with each other for power. Kenya, Nigeria and Zimbabwe are some examples.
28 Conflict incidence is used in the measurement of regime types.
29 Most other indices on ethnolinguistic fractionalisation are time fixed and often based on data from the 1960s. Moreover, they include all ethnographically distinct groups in a country irrespective of whether they engage in political competition.
The PREG index covers 41 of the 47 countries in the panel. Data from Alesina (Citation2003) are used to supplement for Cape Verde, the Comoros, Djibouti, Eritrea, Mauritania and Sao Tome & Principe.
30 Examples of grievances related to oil revenues can be found in the Cabinda Province of Angola, Biafra in Nigeria and the current situation between Sudan and South Sudan.
31 See Table in Appendix 1 for summary statistics.
32 Some countries perform exceptionally well in comparison to the average such as Equatorial Guinea & Gabon, two major oil exporters, Botswana which has large diamond and nickel deposits, Mauritius which has a successful diversified agricultural economy and South Africa which has a relative successful mixed economy.
33 A period of 30 years is fairly short in climatic terms to distinguish clear trends. However, other research has shown that over a longer period of time rainfall in Africa has reduced (see Nicholson (Citation2001); Buhaug (Citation2010); Barrios, Bertinelli, and Strobl (Citation2010)).
34 There has only been a noticeable spike between 2007 and 2009.
35 Civil wars with more than a 1000 battle-related deaths in a given year are relatively rare. In the data-set, only 5 observations are coded for onset of a civil war, 2 of which followed the onset of a minor conflict. In this paper, I focus on intrastate conflict and not interstate conflicts. Between 1981 and 2010, there were only 7 cases of interstate conflict, and only one which lasted more than a year (Ethiopia & Eritrea from 1998 to 2000).
36 See King and Zeng (Citation2001) for an explanation on Rare Event Logit.
37 The number of countries included in the data-set is 47. The first-stage estimations are done for 45 countries due to missing values.
38 I also re-estimated the MSS model using their data-set and changing their rainfall data with the new rainfall estimates. The results for this are in the Appendix 1 in Table .
39 When using time-fixed effects to control for common shocks only current rainfall growth rates are significantly linked to economic growth.
40 The size of the coefficient in columns 4–6 is 0.01 but this is due to rounding as the real coefficient is 0.006.
41 See Table in the Appendix 1 for the first-stage estimations per period.
42 According to Barrios, Bertinelli, and Strobl (Citation2010), hydro-electricity represents about 47% of total power generation in Africa. In a simple regression, I was unable to find a strong link between rainfall and electricity output (results not shown).
43 I used the agriculture and industry value added in constant dollars from the World Bank Development Indicators and calculated a per capita measure to normalise across countries (same approach was used for the service sector).
44 1039 mm versus 993 mm, a margin of only 47 mm.
45 Countries that had a case of conflict onset on average have a growth rate of 1.6%, a 1.9% point difference with the countries that didn’t (3.5%). In terms of anomalies, countries with conflict onset have an average yearly deviation of −0.01 against 0 for the other countries. Countries with conflict onset include 471 observations leaving 918 observations for the other countries.
46 Z-value in case of the rare event logit estimation.
47 This is the case for columns 1, 4 and 6.
48 Data for temperature comes from NOAA for 1981–2008 and is based on gauge stations. The downside of using this data is that it could introduce bias into the model due to measurement error (results not shown). Burke et al. (Citation2009) find that temperature is a stronger predictor of agricultural performance and as well as that temperature has a stronger effect on conflict than rainfall. They focus on civil war incidence though not controlling for past conflicts (Burke et al. Citation2009).
49 Note that in contrast with the studies by Raleigh and Kniveton (Citation2012) and Dunning (Citation2008) the data in this paper are aggregated to the country level instead of the use of disaggregated data which probably causes the differences.
50 MSS found a point estimate of −1.84 for lagged growth rates which wasn’t statistically significant. In their model, estimation current growth rates were negatively linked to conflict onset and significant at the 90% confidence level.
51 Brückner and Ciccone (Citation2010) also found that the results by MSS were not robust to the inclusion of time-fixed effects.
52 It could be that this is related to some of the measurement errors in the economic growth rates but this is no more than a guess.
53 Something that also occurred in the reduced-form OLS regressions (see Table ).
54 As stated in the data, section I uses a different index to measure fractionalisation taking into account the political relevance of ethnic groups and allowing it to vary by decade. This might explain the difference with the findings in the literature.
55 In terms of total employment about 44% works in the primary sector while about 16% works in the secondary sector (data from World Development Indicators for Sub-Saharan Africa 1981–2010).
56 Due to the sparseness of accurate national account data, a more disaggregated approach might be more suitable to identify the effect of different economic sectors on conflict likelihood.
57 Although again a caveat here is that conflict influences economic performance and moreover there is also the mechanical problem of data collection in violence-struck areas.