ArticlePDF Available

Abstract and Figures

Armed conflict and economic growth are inherently coupled; armed conflict substantially reduces economic growth, while economic growth is strongly correlated with a reduction in the propensity of armed conflict. Here, we simulate the incidence of armed conflict and its effect on economic growth simultaneously along the economic pathways defined by the Shared Socioeconomic Pathways (SSPs). We argue that GDP per capita projections through the 21st century currently in use are too optimistic since they disregard the harm to growth caused by conflict. Our analysis indicates that the correction required to account for this is substantial – expected income is 25% lower on average across countries when taking conflict into account. The correction is particularly strong for the more pessimistic SSP3 and SSP4 where expected future incidence of armed conflict is high. There are strong regional patterns with countries with contemporaneous conflicts experiencing much higher conflict burdens and reduced economic growth by the end of century. The implications of this research indicate that today’s most marginalized societies will be substantially more vulnerable to the impact of climate change than indicated by existing income projections.
This content is subject to copyright. Terms and conditions apply.
Environ. Res. Lett. 18 (2023) 024028
10 September 2022
4 January 2023
9 January 2023
27 January 2023
Original Content from
this work may be used
under the terms of the
Creative Commons
Attribution 4.0 licence.
Any further distribution
of this work must
maintain attribution to
the author(s) and the title
of the work, journal
citation and DOI.
The ‘conict trap’ reduces economic growth in the shared
socioeconomic pathways
Kristina Petrova1, Gudlaug Olafsdottir1, Håvard Hegre1,2and Elisabeth A Gilmore2,3,
1Department of Peace and Conflict Research, Uppsala University, Uppsala, Sweden
2Peace Research Institute Oslo, Oslo, Norway
3Department of Civil and Environmental Engineering, Carleton University, Ottawa, Canada
Author to whom any correspondence should be addressed.
Keywords: economic growth, projections, armed conflict, shared socioeconomic pathways (SSPs), climate change
Supplementary material for this article is available online
Armed conflict and economic growth are inherently coupled; armed conflict substantially reduces
economic growth, while economic growth is strongly correlated with a reduction in the propensity
of armed conflict. Here, we simulate the incidence of armed conflict and its effect on economic
growth simultaneously along the economic pathways defined by the shared socioeconomic
pathways (SSPs). We argue that gross domestic product per capita projections through the 21st
century currently in use are too optimistic since they disregard the harm to growth caused by
conflict. Our analysis indicates that the correction required to account for this is
substantial—expected income is 25% lower on average across countries when taking conflict into
account. The correction is particularly strong for the more pessimistic SSP3 and SSP4 where
expected future incidence of armed conflict is high. There are strong regional patterns with
countries with contemporaneous conflicts experiencing much higher conflict burdens and reduced
economic growth by the end of the century. The implications of this research indicate that today’s
most marginalized societies will be substantially more vulnerable to the impact of climate change
than indicated by existing income projections.
1. Introduction
When estimating future socioeconomic scenarios and
their implications, one of the most critical inputs
is the gross domestic product (GDP) and its rate
of growth over the long run (Christensen et al
2018). Extended end-of-century GDP projections are
important in the projection of the impacts and eco-
nomic costs of climate change (Rose et al 2017).
Reflecting the importance of GDP as an indicator of
socio-economic development, GDP projections have
been used to project energy and land use (Popp et al
2017, Riahi et al 2017), food prices (Popp et al 2017),
quality of governance (Andrijevic et al 2020), and
armed conflict (Hegre et al 2016).
Most prominent among the GDP projections in
use is the ENV-Growth model developed by Dellink
et al (2017). This model builds projections of future
economic growth, using a convergence framework
and interacting key long-run drivers of population,
total factor productivity, physical capital, employ-
ment and human capital, and energy and fossil fuel
resources (specifically oil and gas). The projections
are specified as operationalizations of each of the
five shared socioeconomic pathways (SSPs) (O’Neill
et al 2014), and cover the entire 21st century4. Other
projections have also been developed, e.g. Crespo
Cuaresma (2017), Leimbach et al (2017). These are
built on fairly similar assumptions and are highly cor-
related with the Dellink et al (2017) projections5.
Figure 1shows GDP per capita projections
according to the ENV-Growth model for the
4The SSPs were developed by the climate change research com-
munity to harmonize the assumptions that modellers make in
developing projections of the costs of mitigation and adaptation
to climate change. In addition to GDP, the SSPs define alternative
bounding scenarios for variables such as population and education.
5In this article, we only discuss the Dellink et al (2017) since data
for the other projections proved difficult to obtain.
© 2023 The Author(s). Published by IOP Publishing Ltd
Environ. Res. Lett. 18 (2023) 024028 K Petrova et al
Figure 1. ENV-Growth projections for five countries from Dellink et al (2017): Afghanistan, Bangladesh, the Democratic
Republic of Congo, France and Tanzania, SSP2, 2017–2100.
‘middle-of-the-road’ scenario SSP 2 for one high-
income country (France), two lower-middle income
countries (Bangladesh and Tanzania) and two low-
income countries (Afghanistan and Democratic
Republic of Congo). By 2100, the model suggests that
the income of the Democratic Republic of Congo
(DRC) has converged with France, and that the
incomes in Bangladesh and Tanzania will also con-
verge with France within only a few more decades.
Are these projections plausible? Dellink et al
(2017) emphasize that they disregard external shocks
or other non-economic factors affecting productiv-
ity or technological transfer, such as governance or
environmental damages. One such growth-inhibiting
shock is internal armed conflict. Organized polit-
ical violence is often so detrimental to a country’s
economy that it has been termed ‘development in
reverse’ (Collier et al 2003). A number of independ-
ent studies agree that the armed conflicts that his-
torically have afflicted 15%–25% of all countries at
any time, leads to an annual growth shortfall of 2%
per conflict year (Collier 1999, Gates et al 2012)6.
Afghanistan has had continuous armed conflict since
the 1970s (Pettersson and Öberg 2020). Ignoring this
constraint on Afghanistan’s future growth traject-
ory seems unrealistic. Neglecting this feedback means
that we are likely to overestimate future GDP (Buhaug
and Vestby 2019) for Afghanistan, the DRC, and
other conflict-prone countries. Previous efforts that
approximate the risk of armed conflict in the future,
as demonstrated in Hegre et al (2016) and Witmer
et al (2017), suggest that this shortcoming can be
Both armed conflict and economic performance
can interact with climate mitigation and adaptation
efforts (Buhaug and von Uexkull 2021, Gilmore and
Buhaug 2021). Armed conflict curtails economic
6See supplemental information A-1 for a review of these studies.
activity and reduces capacity to development chal-
lenges (e.g. de Groot et al 2022) . Thus, reducing the
burdens of armed conflict is critical for addressing
financial constraints as well as the social and polit-
ical unrest that can hinder efforts to adapt to adverse
climate impacts, especially in the more vulnerable
countries. Mitigation efforts may also be affected by
armed conflict. Furthermore, by reducing institu-
tional capacity, armed conflict could constrain efforts
to coordinate mitigation of greenhouse gas emissions
from national to international levels.
To demonstrate how the dynamics of armed con-
flict and GDP interact over the long-term, we develop
in this article the first joint projections for growth
in GDP per capita and armed conflict that con-
sider the reciprocal effect of the two phenomena on
one another. We use these new GDP pathways to
adjust the ENV-Growth GDP per capita projections
for the plausible losses due to destructive armed con-
flict. To simulate armed conflict and its implications
for GDP, we develop empirical models of the onset
and duration of conflict and the effect of conflict
on GDP growth, as well as a simple model of eco-
nomic growth. We then jointly simulate these out-
comes using the forecasting approach outlined in
Hegre et al (2013,2016). We run the simulation for
each of the five SSP scenarios, and revise the ENV-
Growth model results based on the simulated preval-
ence of armed conflict.
2. Materials and methods
Given their prominence and sophistication, we take
the Dellink et al (2017) projections as our point of
departure, called Yo
it here. We estimate a country-
specific correction dit , and compute a corrected set of
projections as
it =Yo
it dit.(1)
Environ. Res. Lett. 18 (2023) 024028 K Petrova et al
Figure 2. Projections from KC and Lutz (2017). Total global population (a), and average global proportion of population with
secondary education (b). Education levels are unweighted averages of countries’ education level.
Table 1. Fixed-effects OLS results, two growth models, 1960–2016. Detailed specification and results in supplemental information A-2.
Growth model I Growth model II
Coefficient Std. error Coefficient Std. error
Intercept 0.0072 0.014 0.0243 0.014
Conflict 0.0233 0.004 0.0980 0.011
Log education 0.0912 0.027 0.0856 0.027
Population growth 0.5885 0.189 0.6402 0.189
Log population 0.0104 0.008 0.0170 0.008
Int. Population×conflict 0.0260 0.003
Country decay fixed effects Yes Yes
To quantify dit , we estimate a (separate) set of lin-
ear and logistic regression models of the relationships
between economic growth and armed conflict that
reproduce the consensus view on the empirical rela-
tionship between these. We then run two sets of simu-
lations for each of the SSPs: Ic
it, where we jointly simu-
late armed conflict and GDP growth, and Ip
it, where we
simulate GDP growth while ignoring armed conflict.
For each combination of SSP, country, and year, we
calculate the difference dit =Ip
it in simulated log
GDP per capita between each pair of matched simula-
tions. We finally subtract dit from the original Dellink
et al (2017) projections.
Using the Dellink et al (2017) projections as the
base recognizes the many strengths of their approach.
By simply providing the corrections dit, we do not
remake the entire projection process, but rather illus-
trate the need to take armed conflict into account
when thinking about economic growth for the long
2.1. Data
We develop our models of armed intrastate con-
flict with data from the 2017 update of the UCD-
P/PRIO Armed Conflict Dataset (Gleditsch et al
2002, Allansson et al 2017), which records conflicts
between governments and organized armed actors
with a political motivation that lead to at least 25
battle-related deaths in a year. Historic GDP per cap-
ita is derived primarily from the World Develop-
ment Indicators (World Bank 2017)—the same set of
sources as used by Dellink et al (2017).
The exogenous country-level variables in our
model are total population, population growth, and
rates of secondary education attainment, available
from IIASA (KC and Lutz 2017). Figure 2shows
observed and projected total global population under
each of the five SSPs as well as the proportion of
the population that have completed upper second-
ary education. SSP1 (sustainability; green line) and
SSP5 (conventional development; black line) have
optimistic assumptions regarding population growth
and expansion of education. SSP4 (inequality; red
line) assumes minimal expansion of education and
higher population growth, whereas SSP3 (fragmenta-
tion; purple line) have similarly pessimistic education
expansion assumptions and an even stronger popula-
tion growth. SSP2 (blue line) is a middle-of-the-road
2.2. Short-term impact of conflict on growth
Table 1shows the results from a fixed-effects OLS
regression with difference in log GDP per capita from
one year to the other as the dependent variable. Per-
capita growth is higher the higher the education level
of a country, the lower the population growth, and,
most importantly for our purposes, when there is no
conflict7. In model I, log growth is 0.0233 lower in
years when a country experiences conflict, roughly
corresponding to 2.3% lower growth in percentage
7A detailed discussion of the covariates and the estimated coeffi-
cients are found in supplemental information A-2.
Environ. Res. Lett. 18 (2023) 024028 K Petrova et al
terms. In model II, we include an interaction term
between conflict and population size to model that
armed conflict of a given size might be more severe
in smaller countries.
2.3. Simulation procedure
To generate the basis for the conflict-corrected GDP
per capita projections (Ip
it) we expand the ‘dynamic
simulation’ procedure used in Hegre et al (2013,
2016). Explicitly modeling the endogenous connec-
tion between conflict and growth, we simulate both
probabilities of armed intrastate conflict as well as
GDP growth per capita for each year, allowing the
simulations to inform one another. Armed conflict is
a covariate in the growth equation, and growth and
income in the conflict equation. We first estimate the
probability of conflict, and feed that into the growth
models for that year8. The procedure implies estim-
ating a set of underlying statistical models (the one in
table 1as well as a conflict model shown in table A-
3), assuming the projections for population and edu-
cation from IIASA for 2017–2100 (figure 2) are exo-
genous to conflict and growth. The models include a
set of country and region fixed effects, and we assume
these terms are exogenous. Since it would be unreal-
istic that unobserved differences between countries
will remain unchanged over many decades, we reduce
their importance in the future by letting them decay
with a half-life of 20 years as the simulations reach
into the future9.
In a number of repeated simulations, we draw
realizations of model coefficients based on the estim-
ated coefficients and the variance-covariance matrix
for the estimates; calculate probability distributions
for conflict and growth rates for year t0based on
the realized coefficients and the predictor variables,
and randomly draw realized conflict and growth
rates based on these. We then update the values for
the variables measuring historical experience of con-
flict and growth in the country and neighbourhood.
After drawing realized conflict and growth for a year,
we add the simulated growth to the previous year’s
logged GDP per capita to obtain a new value for
the simulated GDP per capita, and repeat for each
year in the forecast period 2017–2100, and record the
8See the supplemental information for details on the modelling
(A-2), on the simulation procedure (A-3), as well as detailed estim-
ation results (A-4).
9The chosen value for the decay of the fixed effects implies a rate
of change in countries’ fundamental social, economic and polit-
ical structures between the faster changes seen in East Asia or oil-
producing Arab countries from the 1960s to today and slower
changes observed in the structures of countries in North Africa or
Latin America. This assumption of slow convergenceis a conservat-
ive assumption. Countries that are currently poor will grow more
rapidly than under a no-convergenceassumpt ionand will therefore
have less conflict, and consequently have a smaller GDP correction.
This also avoids assuming that historical differences are permanent
(no convergence), which is also in line with Dellink et al (2017)
where convergence plays an important part of their model.
simulated outcomes for growth, GDP per capita, and
conflict. The updated conflict, growth, and GDP per
capita variables are then used when simulating the
next year’s values for these three variables. We label
the procedure a ‘dynamic simulation’ since the out-
comes we draw affect the incidence of conflict at time
steps t+2,t+3, etc10.
To even out uncertainty about model specifica-
tions, we run simulations for both sets of growth
models (table A-2) and conflict models (table A-3)
and average over the results. In 40% of the simulations
we have no region fixed effects in the conflict models,
and the remaining 60% are distributed equally over
four different region definitions11. When simulating
conflict, we assume that the underlying unexplained
conflict propensity of the past six years will remain the
same as in the 2011–16 period12. We run 100 simula-
tions for each of the clusters for each of ten imputed
datasets, totalling 5000 simulations, and take the aver-
age of the results to create our corrections.
We calculate dit =Ip
it by running two pairs of
simulations of our economic growth model for the
2017–2100. In the first (Ip
it) we simply assume there
will be no conflicts anywhere, just as in Dellink et al
(2017). In the second (Ic
it), we simulate how much
growth-reducing conflict to expect over the period,
and update the growth paths of countries in which we
simulate conflict using the growth model in table A-2.
The final step in our correction procedure is to
add the difference dit to the original ENV-Growth
model projections to arrive at conflict-corrected
growth projections.
3. Results: corrected GDP per capita
projections 2017–2100
3.1. GDP per capita corrections, global level
Figure 3(a) shows the cumulative difference dit in
(unweighted) global GDP per capita between sim-
ulations where we ignore armed conflict and those
where we take them into account. The corrections
are very large—on average, countries’ GDP per cap-
ita are 20%–30% lower by the century, depending on
the SSP. The conflict trap indeed reduces economic
growth dramatically, and ignoring it is not tenable.
As we show below, for some countries the Dellink
et al (2017) end-of-century projected incomes are 4–
5 times larger than what our more plausible set of
assumptions yields.
Figure 3(b) shows the simulated proportion of
countries in conflict that causes the growth losses
10 See Hegre et al (2013,2016) for further details.
11 The results for the remaining conflict models are reported in
supplemental information A-4.
12 That is, we assume that the ‘temporal fixed effects’ for the 2011–
16 period in table A-3 and the detailed tables in supplemental
information A-4 are the ones guiding the simulations. The underly-
ing conflict propensity was higher in the last decade with data than
in the preceding five-year periods.
Environ. Res. Lett. 18 (2023) 024028 K Petrova et al
Figure 3. Simulation results, global unweighted averages, 2017–2100. SSP1 (green), SSP2 (blue), SSP3 (pink), SSP4 (red), SSP5
requiring this correction. The projected global pro-
portion of countries in conflict is roughly in line
with earlier studies using the same basic setup (Hegre
et al 2016)13. The simulations for SSP1 and SSP5 sug-
gest a clear decline in conflict from current levels, to
less than 15% of all countries at the end of the cen-
tury. This decline is driven by the moderate popula-
tion growth and robust expansion of education under
these scenarios (see figure 2). Conversely, the fore-
casts for SSP3 and 4 suggest an increasing incidence
of conflict (per country, if not per capita), to about
25% of all countries in 2100. This increase is driven
by high population growth and a slow expansion of
education levels.
In any year in the future, then, 10%–25% of all
countries, depending on the SSP, will have an ongo-
ing armed conflict. Our estimates (table A-2) suggest
that every year we simulate that a country is in con-
flict, the country has a growth rate that is on average
2.3% lower than if a similar country avoids conflict.
Over the 84 years of simulation, these growth losses
accumulate, especially in the SSPs where projected
conflict levels are high. In the low-conflict SSP5, the
unweighted average GDP per capita is more than 30%
lower in 2100 than what conflict-ignorant projec-
tions indicate. For the high-conflict SSP3 and SSP4
13 Since we are using more elaborate and credible projections for
education (KC and Lutz 2017) than Hegre et al (2016), the fore-
casts for SSP 3 and 4 are somewhat more optimistic than the pre-
vious study. The forecasts for SSP 1, 2, and 5, on the other hand,
are relatively more pessimistic, since we here include the corrected
growth projections in the conflict forecasts.
scenarios, unweighted average GDP per capita is more
than 35% lower. In these scenarios, the low under-
lying economic growth rate compounds the effect
in a conflict trap. The high population growth and
low education levels suppress income, thus increasing
expected conflict levels, and further decrease growth
rates. Figure 3(c) shows the end-of-century income
correction for all the countries included as a func-
tion of the end-of-century predicted conflict probab-
ility. Countries like the Scandinavian countries have
incomes that are unaffected by conflict. Countries
like Bangladesh (BNG), Tanzania (TZ), Afghanistan
(AFG), and the DRC are predicted to have conflict in
2100 in 40%–60% of the simulations, and our estim-
ated corrections range between 40% and 65%.
Figure 3(d) shows the corrected and uncorrected
ENV-Growth projections for the five SSPs. The ori-
ginal projections (again as unweighted global aver-
ages) are shown as dotted lines. Compared to the pro-
jected increases in income over the next 80 years, our
20%–30% corrections are not large. As we discuss
below, correcting for armed conflict are probably not
sufficient to obtain truly plausible growth estimates
over such a long forecasting horizon, but are clearly a
step in the right direction.
3.2. Region- and country-level results
The cumulative size of the correction in GDP per
capita differs greatly between countries and regions.
Figure 4illustrates the divergence between countries
for the middle-of-the-road scenario (SSP2). The fur-
ther toward the red end of the scale, the larger the
Environ. Res. Lett. 18 (2023) 024028 K Petrova et al
Figure 4. Cumulative correction in SSP2 for GDP per capita by 2100, by country.
Figure 5. Dellink et al (2017), projections (dashed lines) and corrected projections (solid lines), (2017–2100), by region and by
SSP. Top left: East Africa, top right: West Africa, bottom left: South Asia, bottom right: Latin America.
correction. For countries with hardly any correction
to their growth projections, the colour is purple. The
corrections are greatest for countries our model sug-
gests have a high future risk of conflict. This risk is
high for countries with a recent, extended conflict his-
tory, as well as large and poor countries. Mali, Niger,
Afghanistan and Ethiopia tick off many of these
boxes, and the simulated effect of conflict on future
economic growth over time is large. Their future
income levels compared to a peaceful counterfactual
are much more reduced than, for instance, Iceland or
New Zealand that have a very low risk of future con-
flict. Some countries with no recent conflict history,
such as France, Germany and the US, also see decrease
in GDP per capita as a result of projected conflict.
Armed conflict has recently affected a few large, high-
income countries, for example the Basque and North-
ern Ireland conflicts. Spillovers from neighbouring
countries with conflict risks and strong population
growth along some SSPs means that intermittent con-
flict could affect growth in Western countries in the
Inspecting the GDP per capita corrections by
region further illustrates important differences.
Figure 5shows the corrected and uncorrected ENV-
Growth projections for the five SSPs for four regions.
The correction is substantial in East Africa (top left)
for all the five scenarios, reflecting a high frequency of
simulated conflicts in this region. For SSP4, the cor-
rected average GDP per capita in 2100 is under half of
Environ. Res. Lett. 18 (2023) 024028 K Petrova et al
Figure 6. Historical observations and projections, individual countries.
the Dellink et al (2017), original, and the correction
is almost as large for the other SSPs. The adjustments
are less marked in West Africa (top right), which
historically has been more peaceful than its Eastern
neighbours. Likewise, the correction is even smaller in
Latin America (bottom right), a region where armed
conflict is approaching obsolescence14. In South Asia,
on the other hand, our correction is again substantial.
Several countries in the region, e.g. India, Pakistan,
14 Note that the organized criminal violence in the region mostly
falls outside our definition of armed conflict (Allansson et al 2017).
and Myanmar, have had virtually continuous conflict
since independence. Our models suggest that this
will continue for several decades, given their relative
poverty levels and conflict history.
Disaggregating further down to the country level
we return to the illustrative cases shown in figure 1,
to further understand how country-level differences
in input data shape the outcomes.
Figure 6shows corrected and uncorrected GDP
per capita and conflict for Afghanistan, the DRC,
Bangladesh, and Tanzania. The left column shows the
ENV-Growth (Dellink et al 2017) projections as well
as these projections with our correction, for each SSP.
Environ. Res. Lett. 18 (2023) 024028 K Petrova et al
The right column shows the conflict projections for
each SSP.
All four countries are all low- or lower-middle
income and have a high projected probability of
armed conflict. The corrections for Afghanistan and
the DRC have considerable face validity. Afghanistan,
for example, has been continuously at war for forty
years and may conceivably continue to be so for many
decades. If the war takes off 2% annually from the
real growth potential of the country, the loss easily
accumulates to in excess of 80% loss over a century.
The conflict predictions for Bangladesh and Tanzania
are pulled up by their large population sizes and low
income levels. Toward the end of the century, all these
countries are forecasted to see more than 25 deaths in
75% of the years under the pessimistic scenarios SSP 3
and 4. Under the optimistic scenarios, the simulations
yield conflict in about half of the years. These forecasts
seem high, but recall that the population in 2100 is
projected to exceed 200million in both Tanzania and
DRC in SSP 3 and 4, and well over 100million even
in the low-population growth scenario. With a fixed
threshold of 25 deaths in the definition of armed con-
flict, population size is a major predictor of conflict
(Raleigh and Hegre 2009).
The income corrections for these countries are
substantial. In Afghanistan under the pessimistic
scenario of SSP4, Dellink et al (2017) projects an
increase in GDP per capita to about 10 000 dollars by
the end of the century. In the same scenario the pro-
jected income with our correction is slightly below
the current levels of 2000, only a fifth of the ori-
ginal projection. In this scenario, our model predicts
a high probability of continued conflict, given strong
population growth and little expansion of education,
and as such produces severely depressed growth rates.
For the more optimistic scenario SSP5, where Dellink
et al (2017) projects Afghanistan’s income to reach
the implausible value of 140 000 dollars, our correc-
tion still suggests a value about 75% lower. In this
scenario, with lower population growth rates, expan-
sions of education and thus conditions that facilitate
conflict mitigation in the future, the estimated prob-
ability of conflict is far lower than in SSP4. How-
ever, our projections still take into account that the
risk of conflict remains high at first, and this risk is
likely to continue to shape the economic trajectory of
the country for decades to come. The growth projec-
tions for Afghanistan and DRC with our correction
are more realistic than the original for SSP1 and SSP5,
but still likely to be overly optimistic. As the history of
these countries suggest, there are also other sources of
growth failures. The GDP per capita of the DRC, for
instance, fell steadily from 1970 to 1995, for instance,
despite the low levels of conflict in that period.
4. Conclusion
This work improves the understanding of links
between economic development and civil conflict as
well as produces forecasts of future conflict burdens
that are consistent with widely used climate change
scenarios. We successfully model the effect of the
conflict trap on economic growth over the course
of the 21st century, providing a first indication of
how the ENV-Growth projections of GDP per cap-
ita can be corrected for the effect of armed conflict.
Globally, our corrected projections are close to 25%
lower than the original at the end of the century for
the most optimistic Shared Socioeconomic Pathways,
and more than 30% lower in the least optimistic ones.
Thus, the ENV-Growth model (Dellink et al 2017)
clearly over-estimates future growth in conflict-prone
As the ENV-Growth model underlies much of the
existing climate change research, these proposed cor-
rections may have substantial implications for current
estimates of future adaptation and mitigation efforts.
The correction is largest for currently poor and vul-
nerable countries with a conflict history, and suggests
that the resources these societies will have available for
adapting to climate change and other challenges are
much lower than assumed in studies that rely on cur-
rently available projections. These revised GDP pro-
jections that include armed conflict also have implic-
ations for understanding the costs of and capacity for
mitigation efforts. As armed conflict has also been
shown to lead directly to armed conflict in neigh-
bouring countries, these spillover economic effects
may even have adverse effects on an international level
(e.g. providing cover for terrorist activities). Thus,
more importantly, the increase in conflict and result-
ant institutional instability can increase challenges to
the attainment of global agreements and capacity for
climate mitigation policy.
Accounting for the risk of armed conflict is only
one among several issues that remain to be addressed
in economic growth projections. There are several
other governance failures that are less violent but
equally growth-inhibiting, as we noted for the case
of DRC above, and also exemplified by Zimbabwe
and Venezuela. For long-term growth projections to
be realistic, further research should also take broader
governance failures into account. Also, armed con-
flict and other governance failures are likely to affect
other core inputs to growth models. Persistent con-
flicts affect population health, migration patterns,
and undermine education. All of these, in turn, alter
the likely future growth paths of countries. While
these effects will be concentrated in the countries
where the conflict occurs, these effects may also be
Environ. Res. Lett. 18 (2023) 024028 K Petrova et al
experienced regionally in countries that share bor-
ders and more generally, through changes in trade
and other political spillovers.
Data availability statement
The data that support the findings of this study
are openly available at the following URL/DOI:
This material is based upon work supported in part
by the U.S. Army Research Laboratory and the U.S.
Army Research Office via the Minerva Initiative
under Grant No. W911NF-13-1-0307, the MISTRA
Geopolitics programme, Riksbankens Jubileums-
fond programme Societies at Risk, and the European
Research Council Project H2020-ERC-2015-AdG
694640 (ViEWS). The simulations were performed
on resources provided by the Swedish National Infra-
structure for Computing (SNIC) at Uppsala Mul-
tidisciplinary Center for Advanced Computational
Science (UPPMAX). The authors would like to thank
Frederick Hoyles for developing the simulation pro-
gram, Remco Jansen for work on the cluster regions,
Maxine Ria Leis and Hannah Frank for help with
the data, and Chandler Williams for helpful com-
ments. For more information on the ViEWS project
Conflict of interest
The authors have no conflicts of interest to declare.
Ethics statement
All authors have seen and agreed with the contents
of the manuscript and there is no financial interest to
CRediT statement
Kristina Petrova: Validation, formal analysis and the-
ory, data curation, writing, and visualization. Gud-
laug Olafsdottir: Formal analysis and theory, data
curation, writing, and visualization. Håvard Hegre:
Conceptualization, methodology, validation, formal
analysis and theory, writing, project administration,
and funding acquisition. Elisabeth Gilmore: Concep-
tualization, formal analysis and theory, writing, and
funding acquisition.
Kristina Petrova
Gudlaug Olafsdottir
Håvard Hegre
Elisabeth A Gilmore
Allansson M, Melander E and Themnér L 2017 Organized
violence, 1989–2016 J. Peace Res. 54 574–87
Andrijevic M, Crespo Cuaresma J, Muttarak R and
Schleussner C-F 2020 Governance in socioeconomic
pathways and its role for future adaptive capacity Nat.
Sustain. 335–41
Buhaug H and Vestby J 2019 On growth projections in the
shared socioeconomic pathways Glob. Environ. Polit.
19 118–32
Buhaug H and von Uexkull N 2021 Vicious circles: violence,
vulnerability and climate change Annu. Rev. Environ. Resour.
46 545–68
Christensen P, Gillingham K and Nordhaus W 2018 Uncertainty
in forecasts of long-run economic growth Proc. Natl Acad.
Sci. 115 5409–14
Collier P 1999 On the economic consequences of civil war Oxf.
Econ. Pap. 51 168–83
Collier P, Elliot L, Hegre H, Hoeffler A, Reynal-Querol M and
Sambanis N 2003 Breaking the Conflict Trap: Civil War and
Development Policy (Oxford: Oxford University Press)
Crespo Cuaresma J 2017 Income projections for climate change
research: a framework based on human capital dynamics
Glob. Environ. Change 42 226–36
de Groot O J, Bozzoli C, Alamir A and Brück T 2022 The global
economic burden of violent conflict J. Peace Res. 59 259–76
Dellink R, Chateau J, Lanzi E and Magné B 2017 Long-term
economic growth projections in the shared socioeconomic
pathways Glob. Environ. Change 42 200–14
Gates S, Hegre H, Nygård H M and Strand H 2012 Development
consequences of armed conflict World Dev. 40 1713–22
Gilmore E A and Buhaug H 2021 Climate mitigation policies and
the potential pathways to conflict: outlining a research
agenda Wiley Interdiscip. Rev.-Clim. Change 12 e722
Gleditsch N P, Wallensteen P, Eriksson M, Sollenberg M and
Strand H 2002 Armed conflict 1946–2001: a new dataset
J. Peace Res. 39 615–37
Hegre H, Buhaug H, Calvin K V, Nordkvelle J, Waldhoff S T and
Gilmore E 2016 Forecasting civil conflict along the shared
socioeconomic pathways Environ. Res. Lett. 11 054002
Hegre H, Karlsen J, Nygård H M, Strand H and Urdal H 2013
Predicting armed conflict 2010–2050 Int. Stud. Q.
57 250–70
KC S and Lutz W 2017 The human core of the shared
socioeconomic pathways: population scenarios by age, sex
and level of education for all countries to 2100 Glob.
Environ. Change 42 181–92
Leimbach M, Kriegler E, Roming N and Schwanitz J 2017 Future
growth patterns of world regions—a GDP scenario
approach Glob. Environ. Change 42 215–25
O’Neill B C, Kriegler E, Riahi K, Ebi K L, Hallegatte S, Carter T R,
Mathur R and van Vuuren D P 2014 A new scenario
framework for climate change research: the concept of
shared socioeconomic pathways Clim. Change 122 387–400
Pettersson T and Öberg M 2020 Organized violence, 1989–2019 J.
Peace Res. 57 597–613
Popp A et al 2017 Land-use futures in the shared socio-economic
pathways Glob. Environ. Change 42 331–45
Raleigh C and Hegre H 2009 Population, size and civil war.
A geographically disaggregated analysis Polit. Geogr.
28 224–38
Riahi K et al 2017 The shared socioeconomic pathways and their
energy, land use and greenhouse gas emissions implications:
an overview Glob. Environ. Change 42 153–68
Rose S K, Diaz D B and Blanford G J 2017 Understanding
the social cost of carbon: a model diagnostic and
Environ. Res. Lett. 18 (2023) 024028 K Petrova et al
inter-comparison study Clim. Change Econ.
08 1750009
Witmer F D W, Linke A M, O’Loughlin J, Gettelman A and
Laing A 2017 Subnational violent conflict forecasts for
sub-Saharan Africa, 2015–65, using climate-sensitive models
J. Peace Res. 54 175–92
World Bank 2017 World Development Indicators (Washington,
DC: The World Bank)
Full-text available
Little research has been done on projecting long-term conflict risks. Such projections are currently neither included in the development of socioeconomic scenarios or climate change impact assessments nor part of global agenda-setting policy processes. In contrast, in other fields of inquiry, long-term projections and scenario studies are established and relevant for both strategical agenda-setting and applied policies. Although making projections of armed conflict risk in response to climate change is surrounded by uncertainty, there are good reasons to further develop such scenario-based projections. In this perspective article we discuss why quantifying implications of climate change for future armed conflict risk is inherently uncertain, but necessary for shaping sustainable future policy agendas. We argue that both quantitative and qualitative projections can have a purpose in future climate change impact assessments and put out the challenges this poses for future research.
Full-text available
Little research has been done on projecting long-term conflict risks in response to climate change. Such projections are currently neither included in the development of socioeconomic scenarios or climate change impact assessments nor part of global agenda-setting policy processes. In contrast, in other fields of inquiry, long-term projections and scenario studies are established and relevant for both strategical agenda-setting and applied policies. Although making projections of armed conflict risk is surrounded by uncertainty, there are good reasons to further develop scenario projections of the future trajectory of armed conflict. In this perspective article we discuss why making quantitative projections of armed conflict in response to climate change is inherently uncertain, but necessary for shaping sustainable future policy agendas. We argue that both quantitative and qualitative projections can have a purpose in future climate projections and put out the challenges this poses for future research directions.
Full-text available
Climate policies will need to incentivize transformative societal changes if they are to achieve emission reductions consistent with 1.5°C temperature targets. To contribute to efforts for aligning climate policy with broader societal goals, specifically those related to sustainable development, we identify the effects of climate mitigation policy on aspects of socioeconomic development that are known determinants of conflict and evaluate the plausibility and importance of potential pathways to armed conflict and political violence. Conditional on preexisting societal tensions and socioeconomic vulnerabilities, we isolate effects on economic performance, income and livelihood, food and energy prices, and land tenure as most likely to increase conflict risks. Climate policy designs may be critical to moderate these risks as different designs can promote more favorable societal outcomes such as equity and inclusion. Coupling research with careful monitoring and evaluation of the intermediate societal effects at early stages of policy implementation will be a critical part of learning and moderating potential conflict risks. Importantly, better characterizing the future conflict risks under climate policy allows for a more comprehensive comparison to the conflict risk if mitigation is not implemented and graver climate damages are experienced. This article is categorized under: • The Carbon Economy and Climate Mitigation > Benefits of Mitigation Abstract Climate policies that emphasize a fair distribution of benefits and compensation for unintended consequences will moderate the risks of violent conflict and future climate impacts.
Full-text available
This article reports on trends in organized violence, building on new data by the Uppsala Conflict Data Program (UCDP). The defeat of Islamic State (IS) in Syria and Iraq has pushed the number of fatalities, almost 75,600, to its lowest level since the outbreak of the Syrian civil war in 2011. However, this de-escalation in Syria is countered by increased violence in Africa, as IS and other transnational jihadist groups have relocated their efforts there. Furthermore, violence has continued to increase in Afghanistan; UCDP recorded more than 31,200 fatalities in Afghanistan in 2019, which accounts for 40% of all fatalities from organized violence across the globe. The general decline in fatalities from organized violence does not correspond with the trend in the number of active conflicts, which remained on a historically high level. UCDP recorded 54 state-based conflicts in 2019, including seven wars. Twenty-eight state-based conflicts involved IS (Islamic State), al-Qaida or their affiliates. In the past decade, conflicts involving these transnational jihadist groups have driven many of the trends in organized violence.
Full-text available
Weak governance is one of the key obstacles for sustainable development. Undoubtedly, improvement of governance comes with a broad range of co-benefits, including countries’ abilities to respond to pressing global challenges such as climate change. However, beyond the qualitative acknowledgement of its importance, quantifications of future pathways of governance are still lacking. This study provides projections of future governance in line with the Shared Socioeconomic Pathways. We find that under a ‘rocky road’ scenario, 30% of the global population would still live in countries characterized by weak governance in 2050, while under a ‘green road’ scenario, weak governance would be almost entirely overcome over the same time frame. On the basis of pathways for governance, we estimate the adaptive capacity of countries to climate change. Limits to adaptive capacity exist even under optimistic pathways beyond mid-century. Our findings underscore the importance of accounting for governance in assessments of climate change impacts.
Full-text available
The recently developed Shared Socioeconomic Pathways (SSPs) have enabled researchers to explore coupled human–nature dynamics in new and more complex ways. Despite their wide applicability and unquestionable advantage over earlier scenarios, the utility of the SSPs for conducting societal impact assessments is impaired by shortcomings in the underlying economic growth projections. In particular, the assumed economic convergence and absence of major growth disruptions break with historical growth trajectories in the developing world. The consequence is that the SSP portfolio becomes too narrow, with an overly optimistic lower band of growth projections. This is not a trivial concern, since resulting impact assessments are likely to underestimate the full human and material costs of climate change, especially for the poorest and most vulnerable societies. In response, we propose that future quantifications of the SSPs should incorporate the likelihood of growth disruptions, informed by scenarios of the relevant political contexts that historically have been important in curbing growth.
Full-text available
How will local violent conflict patterns in sub‐Saharan Africa evolve until the middle of the 21st century? Africa is recognized as a particularly vulnerable continent to environmental and climate change since a large portion of its population is poor and reliant on rain‐fed agriculture. We use a climate‐sensitive approach to model sub‐Saharan African violence in the past (geolocated to the nearest settlements) and then forecast future violence using socio‐political factors such as population size and political rights (governance), coupled with temperature anomalies. Our baseline model is calibrated using 1° gridded monthly data from 1980‐2012 at a finer spatio‐temporal resolution than existing conflict forecasts. We present multiple forecasts of violence under alternative climate change scenarios (optimistic and current global trajectories), of political rights scenarios (improvement and decline), and population projections (low and high fertility). We evaluate alternate shared socioeconomic pathways (SSPs) by plotting violence forecasts over time and by detailed mapping of recent and future levels of violence by decade. The forecasts indicate that a growing population and rising temperatures will lead to higher levels of violence in sub‐Saharan Africa if political rights do not improve. If political rights continue to improve at the same rate as observed over the last three decades, there is reason for optimism that overall levels of violence will hold steady or even decline in Africa, in spite of projected population increases and rising temperatures.
Climate change threatens core dimensions of human security, including economic prosperity, food availability, and societal stability. In recent years, war-torn regions such as Afghanistan and Yemen have harbored severe humanitarian crises, compounded by climate-related hazards. These cases epitomize the powerful but presently incompletely appreciated links between vulnerability, conflict, and climate-related impacts. In this article, we develop a unified conceptual model of these phenomena by connecting three fields of research that traditionally have had little interaction: ( a) determinants of social vulnerability to climate change, ( b) climatic drivers of armed conflict risk, and ( c) societal impacts of armed conflict. In doing so, we demonstrate how many of the conditions that shape vulnerability to climate change also increase the likelihood of climate–conflict interactions and, furthermore, that impacts from armed conflict aggravate these conditions. The end result may be a vicious circle locking affected societies in a trap of violence, vulnerability, and climate change impacts. Expected final online publication date for the Annual Review of Environment and Resources, Volume 46 is October 2021. Please see for revised estimates.
Significance This study develops estimates of uncertainty in projections of global and regional per-capita economic growth rates through 2100, comparing estimates from expert forecasts and an econometric approach designed to analyze long-run trends and variability. Estimates from both methods indicate substantially higher uncertainty than is assumed in current studies of climate change impacts, damages, and adaptation. Results from this study suggest a greater than 35% probability that emissions concentrations will exceed those assumed in the most severe of the available climate change scenarios (RCP 8.5), illustrating particular importance for understanding extreme outcomes.
The dramatic increase in the number of fatalities in organized violence, seen between 2011 and 2014, did not continue in 2015 and 2016. Rather, the notation of some 131,000 fatalities in 2014 was followed by a steep decline, with just below 119,000 in 2015 and a little over 102,000 fatalities in 2016. Despite the decrease, the number was the fifth highest during the entire 1989–2016 period. Most of the fatalities – over 87,000 – were incurred in state-based conflicts, the main driver behind the trend. Just as the number of fatalities, the number of state-based conflicts, albeit remaining on a high level, continued to decrease in 2016, going from 52 to 49, with 12 of them reaching the level of war, with at least 1,000 battle-related deaths. Also the non-state conflicts dropped in number in 2016, from 73 to 60. This was followed by a decrease in the number of fatalities, and only one conflict caused more than 1,000 deaths. Twenty-one actors were registered in one-sided violence, down by five from 2015. A number this low has only been recorded twice before; in both 2009 and 2010, 21 one-sided actors were listed in UCDP data. The number of fatalities also decreased, going from almost 9,800 to a little over 6,000.
The social cost of carbon (SCC) is a monetary estimate of global climate change damages to society from an additional unit of carbon dioxide (CO2) emissions. SCCs are used to estimate the benefits of CO2 reductions from policies. However, little is known about the modeling underlying the values or the implied societal risks, making SCC estimates difficult to interpret and assess. This study performs the first in-depth examination of SCC modeling using controlled diagnostic experiments that yield detailed intermediate results, allow for direct comparison of individual components of the models, and facilitate evaluation of the individual model SCCs. Specifically, we analyze DICE, FUND, and PAGE and the multimodel approach used by the US Government. Through our component assessments, we trace SCC differences back to intermediate variables and specific features. We find significant variation in component-level behavior between models driven by model-specific structural and implementation elements, some resulting in artificial differences in results. These elements combine to produce model-specific tendencies in climate and damage responses that contribute to differences observed in SCC outcomes - producing PAGE SCC distributions with longer and fatter right tails and higher averages, followed by DICE with more compact distributions and lower averages, and FUND with distributions that include net benefits and the lowest averages. Overall, our analyses reveal fundamental model behavior relevant to many disciplines of climate research, and identify issues with the models, as well as the overall multimodel approach, that need further consideration. With the growing prominence of SCCs in decision-making, ranging from the local-level to international, improved transparency and technical understanding is essential for informed decisions.
In the future, the land system will be facing new intersecting challenges. While food demand, especially for resource-intensive livestock based commodities, is expected to increase, the terrestrial system has large potentials for climate change mitigation through improved agricultural management, providing biomass for bioenergy, and conserving or even enhancing carbon stocks of ecosystems. However, uncertainties in future socio-economic land use drivers may result in very different land-use dynamics and consequences for land-based ecosystem services. This is the first study with a systematic interpretation of the Shared Socio-Economic Pathways (SSPs) in terms of possible land-use changes and their consequences for the agricultural system, food provision and prices as well as greenhouse gas emissions. Therefore, five alternative Integrated Assessment Models with distinctive land-use modules have been used for the translation of the SSP narratives into quantitative projections. The model results reflect the general storylines of the SSPs and indicate a broad range of potential land-use futures with global agricultural land of 4900 mio ha in 2005 decreasing by 810 mio ha until 2100 at the lower (SSP1) and increasing by 1080 mio ha (SSP3) at the upper end. Greenhouse gas emissions from land use and land use change, as a direct outcome of these diverse land-use dynamics, and agricultural production systems differ strongly across SSPs (e.g. cumulative land use change emissions between 2005 and 2100 range from −54 to 402 Gt CO2). The inclusion of land-based mitigation efforts, particularly those in the most ambitious mitigation scenarios, further broadens the range of potential land futures and can strongly affect greenhouse gas dynamics and food prices. In general, it can be concluded that low demand for agricultural commodities, rapid growth in agricultural productivity and globalized trade, all most pronounced in a SSP1 world, have the potential to enhance the extent of natural ecosystems, lead to lowest greenhouse gas emissions from the land system and decrease food prices over time. The SSP-based land use pathways presented in this paper aim at supporting future climate research and provide the basis for further regional integrated assessments, biodiversity research and climate impact analysis.