The Political Economy of "Natural" Disasters

The Political Economy of
“Natural” Disasters

Charles Cohen
Eric Werker




Working Paper


08-040

Copyright © 2008 by Charles Cohen and Eric Werker
Working papers are in draft form. This working paper is distributed for purposes of comment and
discussion only. It may not be reproduced without permission of the copyright holder. Copies of working
papers are available from the author.

---

The Political Economy of “Natural” Disasters



Charles Cohen
Vice President, Sankaty Advisors, Boston MA

Eric Werker
Assistant Professor, Harvard Business School, Boston MA

2008

ABSTRACT


Natural disasters occur in a political space. Although events beyond our control may
trigger a disaster, the level of government preparedness and response greatly determines
the extent of suffering incurred by the affected population. We use a political economy
model of disaster prevention, supported by case studies and preliminary empirics to
explain why some governments prepare well for disasters and others do not. We show
how the presence of international aid distorts this choice and increases the chance that
governments will under-invest. Policy suggestions that may alleviate this problem are
discussed.




Key Words: Natural Disasters; Humanitarian Aid; Samaritan’s Dilemma.

---

I.
INTRODUCTION
Natural disasters occur in a political space. They are not driven by politics, nor are they
immune from politics. Incentives faced by human actors can affect the prevention, mitigation,
and damage of natural disasters, even if they cannot affect the likelihood of rainfall in a specific
area or seismic activity along a particular fault line. This is hardly controversial. The vast
literature on disaster prevention and response has appreciated the political dimension of disasters
for decades (Olson, 2000; Platt, 1999; Blaikie et al, 1994; Albala-Bertrand, 1993; Bommer,
1985; Cuny, 1983; Davis and Seitz, 1982; Diggins, Wright, and Rossi, 1979; Abney and Hill,
1966). There is empirical evidence of the relationship: in the United States, political
considerations may explain half of all federal disaster relief (Garrett and Sobel, 2003) and
electoral factors influence whether a president issues a disaster declaration (Reeves, 2005);
around the world, disasters tend to be more severe in poorer countries that are poorly run
(Stromberg, 2007; Kahn, 2005).
This paper represents the first attempt to synthesize these observations into a formal
model of disaster mitigation that contains empirical predictions on the severity of disasters. In
addition, this paper is the first to systematically incorporate the political aspect of disaster in the
heart of the model and recognize the feedback between policy interventions and the seriousness
of the disasters themselves. By bringing free relief to countries afflicted by so-called natural
disasters, international humanitarians can create problems of moral hazard with a time-
inconsistency nature. In the event of a government being ill prepared for a disaster, international
relief effectively rewards bad behavior on the part of the poor countries’ governments.
We adopt a precise, mathematical definition of disaster when the model is introduced in
Section II.1 Our definition is in spirit with Erikson (1976): “[Disasters] involve considerable

2

---

harm to the physical and social environment; they happen suddenly or are socially defined as
having reached one or more acute stages; and something can be done to mitigate their effects
before or after they happen” (quoted in Kreps, 1998: 33). In other words, disasters involve a
stochastic negative shock whose severity can be affected through a process of prevention and
relief. It is precisely that process that we model. Henceforth, “shock” refers to the natural act
itself—the volcanic eruption, earthquake, drought, etc.—and “disaster” refers to the net impact
of the shock on the population.
Natural disasters have killed over 62 million people world-wide since 1900
(OFDA/CRED 2003). This is approximately the same number as all those killed in both World
Wars,2 yet scarce attention has been paid to natural disasters in the economics and political
science literature, while dozens of articles on conflict are published each year. Over 85 percent
of the deaths occurred between 1900 and 1950, and a little over one million deaths from natural
disasters have occurred since 1990.3 Certainly part of the credit for the relatively small number
of disaster deaths in the last decade is due to the efforts of the global humanitarian community,
and to its ever-increasing resources and effectiveness. In the month of October 2005 alone,
international humanitarians and national Red Cross chapters responded to natural disasters in the
Central African Republic, Costa Rica, El Salvador, Guatemala, India, Indonesia, Mexico,
Nicaragua, Pakistan, Paraguay, Romania, and the Sudan (Reliefweb 2005).
In this paper, we take as given that the relevant actors work in a political space, and we
model that space using a reduced-form framework. National governments care about the social
welfare of their citizens, but also want to maximize government income. These governments
can spend on both preventative and palliative measures to lessen the impact of a potential natural
shock. International humanitarian organizations can step in with ex-post relief to poor countries

3

---

that experience natural disasters. This setup produces a number of important results, some of
which have been independently noted in different strands of the literature.
The model predicts that rich governments and governments that care about social welfare
spend more on disaster prevention and mitigation. Governments can use natural disasters to
redistribute power through the political effect, favoring disaster spending in regions that are
politically aligned with the party in power. Governments are less prone than local populations to
“insure” against disasters through preventative measures, because a regional shock has less
impact on the national government’s income than on that of the local region.
The addition of humanitarian aid to the model produces a bailout effect: governments
under-invest in disaster prevention when they know that they will be bailed out in the event of
disaster. This effect is mitigated for pariah states, which may not have access to international
aid. In the extreme, we can witness a racket effect, where governments can deliberately neglect a
population so as to attract—and steal—humanitarian aid in the event of a disaster. Governments
without other sources of external income are more likely to be influenced by the racket effect. In
addition, in the case of shocks that decimate local populations, international organizations will
tolerate higher levels of theft to deliver urgently needed aid. This can lead to a desperation
effect: in dire circumstances, rapacious governments have a stronger ability to increase their level
of theft.
These results have policy implications for reducing the severity of natural disasters.
First, the international humanitarian community must be involved in disaster prevention if it is to
offer free relief. Second, whenever possible, disaster relief should be provided locally so as to
reduce the significance of the racket and political effects on the central government. Third,
political development, in the form of more responsive governments and less intrastate conflict,

4

---

will reduce the severity of disasters. Fourth, in particularly problematic areas, governments can
be given extra payments for proper disaster preparation, including establishing and enforcing
efficient regulations.
This paper is related to two large bodies of research in political science: political
economy of war and foreign aid. In essence, disasters can be used as a blunt policy instrument to
target or reward populations and to enrich a government, and they can be mitigated with outside
effort. Corollary works on the political economy of war include investigations into why the state
might target its own population (Valentino, Huth, and Balch-Lindsay, 2004; Harff, 2003; Krain,
1997), and the effectiveness of benevolent outside intervention on internal conflict (Doyle and
Sambanis, 2000; Paris, 1997; Diehl, Reifschneider, and Hensel, 1996). By highlighting the
potential of foreign assistance to affect domestic outcomes we are adding to a rich literature on
the distortionary impact of foreign aid that has hitherto focused on non-disaster aid (Goldsmith,
2001; Bräutigam, 2000; Lancaster, 1999; Uvin, 1998).
The rest of the paper is organized as follows. In Section II, the model is presented and
the terms are defined. Section III solves the model in the case of no outside humanitarian relief,
while Section IV allows for external disaster aid. Policy implications and conclusions are
offered in Section V.

II.
A POLITICAL MODEL OF DISASTERS

Before making any predictions on the severity of disasters we set up the model from
which the results will be derived.
A. Disasters, Prevention, and Relief

5

---

Assume that a natural shock, such as an earthquake or hurricane, strikes with an
exogenous probability q, and causes monetary damage k. We assume that governments cannot
control the probability of such a shock, only the shock’s impact. We model the impact of the
shock as a per capita loss k − f (b, a) , where f(b,a) is the total level of relief from preventative
spending done before the shock hits (b) and palliative spending done after the shock hits (a). We
assume that f is concave. Preventative and palliative spending may either be complements (that
is, a dollar spent on prevention increases the benefit of a dollar spent on relief, e.g. a dam) or
substitutes (spending on prevention decreases the per-dollar benefit of spending on relief, e.g.
famine relief), and this will play a key role in our analysis. In addition, f has a maximum value
of k (disaster aid cannot make improvements that bring utility above the non-disaster state).
We define a disaster as the net impact of a shock, or k – f(b,a). The initial disaster is
given by k – f(b,0); the level of damage reported after the shock first hits. In other words, a
shock k’s immediate impact on a population is governed by the amount of disaster prevention b
that a government has undertaken. While the net impact of the shock is further determined by
any palliative care a, the disaster that makes the newspaper will not have yet been attenuated by
the relief activity. For the most part, the distinction between a “disaster” and an “initial disaster”
is helpful only in explaining the observed level of disasters around the world. For the purposes
of government decision making, f(b,a) and not f(b,0) is most important.
B. Government Spending
The government may either spend money on disaster prevention or allocate it to other
uses. The government derives utility from two sources: social welfare from disaster prevention
and income that it uses for other purposes. Formally, the government has a utility function
v( ,
w I ) , where w is social welfare of the population and I is government income. An individual

6

---

citizen’s welfare is given by a utility function u(y), where y is an exogenous level of individual
income. If a shock hits, a citizen’s income is reduced by k - f(b,a), and hence her utility drops to
u(y – (k - f(b,a))). Following a representative consumer model, we assume
v(w, I ) = ∑λ u
+
.
i
(yi ) v(I)
i
In words, the government is a Benthamite social planner, assigning different weights λ
i
to different regions (the regions are indexed by i). This λ reflects the government’s inherent
i
interest in region i. We might expect populous regions, swing regions, or politically organized
regions to have high values of λ . In racially or ethnically polarized countries, λ may even be
i
negative in some regions. ∑λ u y is simply the weighted average of citizen utility across the
i
( i )
i
country. The government also cares about income, through a concave function v(I ). We are
assuming that government utility is separable in income and social welfare. If this is not true,
then we need to worry about the complementarity or substitutability of these inputs. We present
this simpler version that captures all the essential insights of our model without unnecessary
technical complications.
This is not a general equilibrium model; we have decided to ignore issues regarding
taxation or supply-side responses to government actions, as these are not central issues in our
paper. In particular, we ignore the fact that government income is actually a function of the
income of the citizens of the country.
The form of this utility function should not necessarily be interpreted literally as the way
that central governments make decisions, but rather as a reduced form for a series of vastly more
complicated decisions. We also recognize that central government decisions are not perfectly
coordinated;4 however, we still think that this functional form will capture the most important

7

---

dynamics. Even though the government may not have preferences directly over these inputs, it
will behave as though it does, and hence this utility function serves as a consistent representation
of these preferences.5 In addition, one can also interpret our formulation as the reduced form of
a more explicitly political model.6 We are not as interested in the subtle dynamics of such a
model as with the interaction of national governments with international relief organizations, and
hence we will use this reduced form throughout the paper.
We abstract from the much larger spending decisions that the central government must
make and consider only that the government has a fixed supply of money I to spend on disaster
prevention or to put aside for other spending. In this model, each country is made up of n
distinct regions to which funds can be allocated, which we index by i. The government spends b i
on prevention and a on relief in region
i
i.

III. GOVERNMENT SPENDING AND DISASTERS IN THE ABSENCE OF AID
In this section, we solve the model without strategic interactions with an outside relief
agency, derive first-order conditions, and take several comparative statics.
A. Efficient Disaster Spending
As a reference case we assume that the government is simply trying to maximize the
utility of one representative citizen. Assume that the unit cost of preventative measures is p and
b
of palliative measures p . Hence, the government is solving the problem (recall that
a
q is the
probability of a disaster):
max (1 − q )u (y − p b +
−
−
,
−
−

b
) qu (y (k f (b a)) p b p a
b
a
)
b , a

8

---

However, note that we can write this in terms of two variables of greater interest: the total level
of relief r=f(b,a), and the level of prevention b. These two variables implicitly define the level
of palliative spending:
f (a(r,b),b) = r .
In words, a(r,b)is the level of palliative aid that must be done to achieve a total level of relief r,
given that an amount b of prevention was provided. This implies:
a
∂ = 1
r
∂
f a

a
∂
− fb
=
b
∂
f a
In addition, the total cost for this spending is now given by a function c(r,b), defined by:
c(r,b) = p a ,
+
.
a
(r b) p b
b
Taking the derivative of this with respect to r:
∂
∂
1
. c
a
= p
= p

(1)
a
a
r
∂
r
∂
f a
c
∂
a
∂
fb
= p
+ p = p − p
(2)
a
b
b
a
b
∂
b
∂
f a
Hence our maximization problem is now:
max(1− q)u(y − p b +
− − −
,
(3)
b
) qu(y (k r) c(r b)
r ,b
Taking the derivative of (3) with respect to r gives us the first of our two first order conditions:
⎛
∂c ⎞
u'(y − (k − r) − c(r,b) ⎜1 −
⎟ = 0
⎝
∂r ⎠
Substituting in from equation (2) yields:
1
1− p
= 0
a
f a

9

---