natural disasters, mass media and political accountability
TRANSCRIPT
Natural Disasters, Mass Media and
Political Accountability
Brian Albrecht, Giulio Bonazzi, Laura Hopkins
June 13, 2014
Abstract
A growing field has developed that examines the interplay between
natural disaster, politicians, and media. Less work has been done to
examine the role of politicians in influencing the media during times of
natural disasters. The decisions made by politicians can greatly affect
the extent of damage from a disaster. The media can influence the
voters with information about the politician’s actions. This creates
an incentive for the politician to provide support before and after a
disaster. With a simple model, we show that under certain conditions
an equilibrium exists in which an opportunistic politician will (1) exert
effort to help a disaster, (2) bribe the media to cover up the damage,
and (3) do nothing. All three with occur with positive probability.
1
“Whoever controls the media,
controls the mind”
Jim Morrison
1 Introduction
In the summer of 2010 with large parts of Russian forests up in the flames,
the Russian media reportedly failed to provide real-time updates of massive
wildfires (Bovt 2010). Close to 56,000 people in Russia died as a result of the
major heat-wave and wildfires (Guha-Sapir et al. 2011, 15). 35 people died
directly from the wildfires and around 800 homes burned down (Williams
et al. 2011, 6). One estimate puts the damage at 1.5% of Russian GDP
(US$3.6 billion), including roughly 1/3rd of wheat production is included
(Porfiriev 2014, 92). Other estimates, which neglect the loss of crops, put
the estimate at 0.2% of GDP (ibid.).
2010 was not the only time in recent history that major wildfires torched
Russia. In 2003, 2008, and 2012, wildfires covered more area than in 2010
(Shvidenko and Schepaschenko 2013, 685). In these case, there has been no
suggestion of media distortion. However, each year was different and hard to
compare. As Quiroz-Flores and Smith note (2012), disasters never come with
a counter-factual and rarely even have a comparable event. No two events
are the same and- besides earthquakes- there is no universal standard for
measuring the level of a disaster. Even the Richter scale is far from perfect
in comparing earthquakes, as the direction and duration of the tremor are
crucial to understanding their impact. This lack of comparability from one
2
disaster to the next is an obstacle to understanding the impact of govern-
mental effort on disaster mitigation.
The relationship between natural disasters and governments is subtle.
Initial actions of governments can greatly impact the extent of a a disaster.
On the one hand, governments can act to prevent the impact of a disaster,
should one occur (ex-ante). On the other hand, the government can respond
to the situation following a disaster with palliative treatment (ex-post). This
endogeneity creates further complications for researchers attempting to un-
cover the actions of governments.
If the government captures the media to suppress reports of a natural
disaster, evidence will clearly be hard to come by. We have collected some
anecdotal evidence where governments had an interest in changing the flow
of news, through full media capture or modification. Although these present
different contexts and the extent of the media capture changes, they all have
a common pattern: preventing the truth from reaching the citizens.
Bearing in mind the difficulties in comparing one disaster to another,
we consider the differences in the flows of information that Chinese media
passed during the 2003 Sever Acute Respiritory Syndrome (SARS) epidemic
and the Sichuan earthquake in 2008. The Chinese media is assumed to be
heavily captured. In the case of the SARS outbreak, the Chinese government
limited the flow of information to avoid negative economic consequences and
internal instability (Zhang 2003).
In 2008 a massive earthquake hit the region of Sichuan. The Beijing’s Central
Propaganda Department has then decided to transmit only proregime news.
This was mainly because the incompetence of the local government was the
3
cause of schools and buldings collapse (John J. Tkacik 2008) .
In 2011, an earthquake in Japan led to severe nuclear disaster. The
Japanese government quickly moved to prevent journalists from accessing the
area close to the reactor, and raised the legally acceptable level of exposure to
radiation. Further, the government introduced a new law that expanded the
definition of official secrets to cover the nuclear disaster, preventing public
access to information (WashingtonBlog 2013).
Whilst these stories do not provide evidence of collusion between media
and the government in the event of a natural disaster, they do raise ques-
tions. When will the government exert effort to prevent or respond to a
natural disaster? What trade-off does the government face between these
two actions?
This paper develops a model for analyzing such questions. We build on
two areas of research which social scientists have heavily analyzed in the past:
(1) the role of institutions in alleviating the damage of disasters and (2) the
role of media in influencing policies. In the past decade or so, a literature
has started to join these two; explaining the role of media in influencing
policies to deal with a disaster. Political scientists have been the driving
figures behind this analysis of media and disaster policies, often looking at
cross-country surveys. Economists have also contributed with more formal
models. We follow this recent economics work and build a simple model for
understanding the different factors at play between politicians, the media,
and voters before and after a disaster strikes (Besley and Burgess 2002; Besley
and Prat 2006; Choe and Raschky 2011).
4
2 Institutions, Natural Disasters and Media
Going back to Amartya Sen’s work, economists and political scientists have
studied the effect of political institutions on natural disaster relief (Sen 1981).
Political institutions can cut disaster casualties, sometimes greatly (Cavallo
and Noy 2009, 13). “Institutions play a role in shielding the population from
natural-disaster death,” even though countries with better institutions are
not less likely to experience disasters in the first place (Kahn 2005, 283).
There is good reason to believe that governments want to limit the dam-
age from natural disasters. Citizens attempt to hold politicians accountable
when disaster strikes, even for things that are not controlled by the politi-
cians, such as natural disasters or shark attacks (Achen and Bartels 2004).
Cross-country analysis suggests that natural disasters increase instability, es-
pecially among authoritarian regimes (Quiroz-Flores and Smith 2012). Even
political regimes like Russia’s under Putin and Medvedev could feel this pres-
sure. This does not only occur in authoritarian regimes. In the United States,
voters punish elected officials at all levels over natural disasters (Malhotra
and Kuo 2009; Gasper and Reeves 2011).
It is also largely recognized the important role that media play in gov-
ernment accountability. Whether or not the media reports on an event can
have a great impact on the amount of aid spent. For example, disasters that
occur during major newsworthy events, such as the Olympics, receive less
aid in the United States than events that are not competing for news atten-
tion. The causation likely does not come from the Olympics. Clearly, media
brings attention to disasters which directly or indirectly influences politicians
5
(Eisensee and Stromberg 2007).
Besley and Burgess show that governments are more responsive to falls
in food production and flood damage in areas with higher newspaper cir-
culation (2002). In their model, voters learn about politicians through the
media. Since the media can amplify the signal that a politician is sending to
their ultimate target, voters, the media can encourage more support after a
disaster.
An emerging literature has developed theoretical models to investigate
how media influence political outcomes. The mechanism according to which
media influence voters is very simple: if, to vote, people look for information
about politicians, then media, by providing information, might be able to
influence the voting decision (Stromberg 2007). A more informed electorate
leads to a more responsive government (Besley and Burgess 2002).
Clearly, a responsible media encourages politicians to work, help voters,
and limit the disaster damage. However, that is not the only direction of
influence. If politicians cannot easily limit the disaster, maybe they can
limit the amount that people know about it. The most obvious way to limit
knowledge is by restricting media coverage. What the citizens do not know
cannot hurt the politician. A restricted media might be a good thing for a
politician.
The media helps influence policies. This saves lives by procuring resources
after a disaster has struck. While disaster relief does increase with the sever-
ity, other factors, such as media coverage, drive a large part of the relief
efforts (Stromberg 2007, 221). Therefore, when analyzing political actions
related to natural disasters, it is important to recognize the role of media. A
6
restricted media hurts citizens and voters.
For example, Besley and Prat develop a model where politicians can
choose to “capture” the media (2006). If newspapers are simple profit-
maximizers, any bribe equal to the profits a media outlet could earn as a
monopoly will result in the media withholding information. Because any
one firm can break the news, capture is expensive even where it is possible.
In Peru, for example, the media cost roughly 100 times as much as judges
and politicians to bribe (McMillan and Zoido 2004). This capture manipu-
lates the media reports and ensures that bad news about a politician does
not reach voters. Politicians can easily manipulate the news if the media
is not independent. Media capture is not a recent phenomenon. Empirical
evidence show that the relationship between media and government char-
acterizes many countries, such as Italy, Mexico, Thailand (Besley and Prat
2006).
While bribing is the clearest example of media manipulation, politicians
can do more. Using data from fieldwork in Romania, Leeson and Coyne show
the many ways that politicians can manipulate media reports (Leeson and
Coyne 2005). While direct control through ownership is the most obvious
way, Leeson and Coyne point out that the government owns much of the
infrastructure that media outlets rely on. Other forms of pressure include
financial pressure through taxes and ad-revenue and control over entry into
the media market. Clearly, thinking of media manipulation as a result of
only bribes or state-ownership leaves something out. Politicians have many
levers of varying impact to manipulate the information that reaches voters.
Since a politician ultimately cares about voters, it is reasonable to believe he
7
will use those levers if he has incentives too.
When do politicians have the incentive to manipulate the media? This
paper develops a simple model to analyze this relationship between media,
political institutions, and disaster relief.
8
3 Model Setup
We use a retrospective voting model similar to the one used in Besley and
Prat (2006). As in their paper, we develop a model that includes a politician,
P , a media outlet, M , and a homogeneous set of voters, V . Unlike Besley and
Prat, however, the politician in our model is opportunistic and not selected
from a set“good” and “bad” politicians.
Our aim is to discover a setting of complete information in which there
exists an equilibrium whereby the politician mixes across different actions,
sometimes doing the “good” thing and sometimes the “bad”, and sometimes
being “lazy”.
The politician has three actions. He can do nothing about the disaster
(“lazy politician”). He can take costly preemptive actions to reduce the
likelihood of a disaster (“good politician”). He can bribe the media to cover
up a disaster (“bad/ corrupt politician”). The media has two actions. It can
report there has been a natural disaster or not report. The voters also have
two actions. They can reelect the incumbent politician or they can vote in a
challenger.
An extended form representation of the model is included on pages 22
and 23 for reference. The timing is such that:
1. The incumbent takes an action, Ap ∈ {N,E, T}:
• N : the politician does nothing; the cost of N is 0;
• E: the politician exerts effort to reduce the likelihood of a natural
disaster; cost of E is e;
• T : the politician offers a bribe to the media; cost of T is t.
9
2. In the event a bribe is offered, the media chooses to accept or reject it.
3. Nature creates a natural disaster with probability d, unless the politi-
cian took preemptive action. In that case, the probability is s, such
that s < d.
4. The media reports that there has been a natural disaster or does not
report a natural disaster, AM ∈ {θ, ∅}.
5. The voters observe the message of the media and vote for the incumbent
or the challenger, AV ∈ {I, C}.
6. Payoffs are realized.
While we model the politician’s action, E, as reducing the probability of
a disaster, it could easily be interpreted as the probability that any single
person is affected by the disaster. This would be more in line with the
examples from earlier where politicians can reduce the damage of a disaster
and not the actual disaster itself.
Voters receive all of their information from the media. They cannot tell
the difference between no report because the media accepted a bribed and
no report because the natural disaster did not occur. When they see a report
of a disaster, they are not sure that the politician exerted effort to reduce
the likelihood of this or not. We denote the information set of no report as
empty, ∅, and that of the report of a disaster as θ.
The media receives an exogenous benefit from reporting the truth, `. This
could come from future loss of readership, since the truth is revealed when
payoffs are realized. It could also be a loss from an intrinsic motivation for
honest reporting. Either way, we treat it as exogenous.
10
We assume that the accepted bribe is credible. If the media reports
the disaster after accepting the bribe, the bribe is revoked. The reverse
commitment is also true, such that the politician is committed to paying the
bribe if the media withholds reporting the disaster.
The politician gains rents, R, from being in office. Where the challenger is
elected, he receives 0. The media has revenues ` when they print the correct
news story, and 0 otherwise. The voter receives g utility from the incumbent
where he took preemptive action i.e. a ”good” politician; b where he was
”lazy” or ”bad”; and x from a challenger, such that g > x > b. When a
natural disaster occurs, the voter suffers damages D.
Since the setting is retrospective, payoffs are not realized at the time of
the reelection decision. The payoffs are observed in the period following the
end of the game.
The model has two parts. The first is a bargaining game between the
politician and the media. The second part is the voting game. Together,
these two parts determine the equilibrium of the overall game.
Ours is not a game of adverse selection. It is a game of complete but
imperfect information, where the voters are limited in their ability to identify
the actions of the politician.
The full extensive form of the game is included in the Appendix for clarity.
11
4 Finding Equilibria
Elimation of dominated strategies
Proposition 1: The politician never offers a bribe that the media
rejects
The media receives revenue ` from reporting the news of a disaster, and zero
otherwise. The media can choose to accept or reject a bribe when offered.
In equilibrium, the politician induces acceptance by offering a bribe of an
amount to cover the media’s potential revenue ` from reporting. Since the
politician moves first and can choose an amount to bribe, he will either offer
no bribe or the lowest amount possible to capture the media. In our model,
this is simply `. This minimum amount could be endogenous. For example, it
could be a function of the number of media outlets, expected media revenue,
and transactions costs between the media and politician (Besley and Prat
2006).
The transfer is executed ex-post, such that the media has a dominant
strategy to suppress a signal of a natural disaster, once the bribe has been
accepted.
Proposition 2: The media always reports the truth if it has not
been captured.
Since the media always has perfect and complete information, it knows at
which decision node it is. Therefore, the media never has to guess. Since the
media gets utility from reporting the truth from future revenues or prestige,
12
it has a dominant strategy to report when a it observes a signal of the natural
disaster and has not been bribed. As stated above, if the politician offers a
bribe of ` the media will accept it and suppress any report of the disaster.
Taking Proposition 1 and 2 into account, the media has a dominant strategy.
The media should accept the bribe and report nothing if the bribe is at least
as large as its lost revenue (t ≥ `). The media should report the truth if the
bribe is less than its lost revenue t < `. This simple point eliminates many
of the possible branches from any equilibrium strategy.
Pure strategy equilibria
Since the media has a dominant strategy of accepting the bribe if it is at least
as large as `, and otherwise reporting a disaster if it strikes, and reporting
nothing if it does not, the game reduces to play between the voters and the
politician. The media simply adds a noise into the system. Now the voters
don’t know whether a disaster strikes because the politician did nothing or
did something, but there was not a disaster. Similarly, if there is no report,
the voters don’t know if it was because the politician did nothing, exerted
effort, or bribed the media. Thinking about the game in these terms makes
the equilibrium strategies more obvious.
Proposition 3: No equilibrium exist where the politician always
bribes the media.
Contrary to Besley and Pratt (2006), our model does not include an equi-
librium where the politician always bribes the media. If the media is always
13
bribing the media, the voter is best off always voting for the challenger, no
matter what his signal is. Since the politician incurs a cost in bribing, he is
better off doing nothing. The difference between our model and theirs comes
from our assumptions about information. Their game involved incomplete
information where the voters didn’t know the extent of the bribing. Since
voters in our model have complete information, in equilibrium they know the
proportion of the time the media is being bribed.
The real world is somewhere in between our model and theirs. Unless the
required bribe amount is zero, it is unreasonable to believe the media will
always be bribed. Since the voters will know that the media is always bribed,
the politician cannot persuade them with the media. The media provides no
information to the voter and loses its effectiveness. The politician must only
bribe the media up to a certain amount, if he does at all, so that the media
remains credible and useful to the voters.
Proposition 4: No equilibrium exists where the voters always re-
elect the incumbent.
Suppose the cost of the action E is quite low. At first it might seem reason-
able that the politician will always do E and the voters will always vote for
the incumbent. However, if the voters always vote for the incumbent, even
if they receive news of a disaster, the incumbent has an incentive to deviate.
As long as the cost of effort is above zero, the politician will be better off
doing nothing and still winning the election. Therefore, the voters’ potential
strategy {I, I} cannot be part of an equilibrium.
This makes sense. If the voters vote for the incumbent no matter what,
14
they lose their chance to act as a check on the politicians. By sometimes
voting for the challenger, the voters can provide incentives for the politician
to exert effort, assuming the effort is not too costly. Again, this is contrary to
Besley and Pratt (2006). In their model, equilibria exist where the politician
always bribe the media and the voters always reelect the incumbent.
Proposition 5: No equilibrium exists where the politician always
exerts effort.
If the politician is always exerting effort, the voters have an incentive to
vote for him. This exists even if they know a disaster struck. They prefer a
politician who put in effort, even if the disaster struck. This leads the voters
to reelect, no matter what their signal is. However, as we showed above, this
creates an incentive for the politician to do nothing, since the voters vote for
him anyway.
Mixed strategy equilibrium
Although certain pure strategies are ruled out, we look for the possibility of a
mixed strategy equilibrium. Such an equilibrium allows for different actions
to be played from both players. With a certain probability the politician
will do each action a certain probability. Also, the voter will vote for the
incumbent with a certain probability, given his information from the media.
A mix strategy captures some of the
The voter receives information from the media. The voter will receive
information regarding a natural disaster θ, or receive no signal ∅. The voter
15
mixes his strategy for voting in the incumbent politician at each information
set.
The equilibrium conditions for which the voter is indifferent between re-
electing the incumbent politician and voting for the challenger are 1:
πe =x− bg − b
πt =(d− s)(g − x)
d(g − b)
For the voter to play a mixed strategy, it must be that the probability
that the incumbent exerted effort πe, offsets the uncertainty faced by the
voter. Specifically, πe must be equal to the ratio of the reduction in utility
caused by voting in a lazy politician, compared with a challenger or good
politician.
The politician plays a mixed strategy to equate his expected payoffs under
each of his actions.
Let y be the probability a voter votes for the incumbent given a report of
a natural disaster θ, and z be the probability a votes votes for the incumbent
give no report ∅.
Then, the expected utility of the politician from each action:
N : dyR + (1 + d)zR
E : sy(R) + (1− s)z(R)− e
T : zR− t
The equilibrium conditions for which the politician is indifferent between
1Details in the appendix
16
his three possible actions are 2:
y = z − e
R(d− s)
y = z − t
dR
y = z +e− tsR
which yields the following property of an equilibrium in which the politi-
cian plays a mixed strategy:
e
t=d− sd
For a mixed strategy equilibrium to exist for the politician, the ratio of
the relative costs between taking action E and action T ( et) must be equal
to the percentage change in probability of a disaster from taking action E.
In the extreme, where t = e, we see that for a mixed strategy equilibrium to
exist, action E must lead to a 0 probability of disaster.
In this mixed strategy equilibrium, voters provide a check so that politi-
cians sometimes do effort. However, their imperfect information prevents
the voters from always pressuring the politician. This equilibrium allows for
voters who know that the media is sometimes untruthful, yet nevertheless
still sometimes vote for the incumbent.
2Details in the appendix
17
5 Extensions
Our simple model highlights some important parts that are not included in
others. Since we have a mixed strategy equilibrium, our model allows for
probabilistic actions from voters and politicians. This is not an option in
other models (Besley and Prat 2006).
However, our simple model is lacking in some ways and could be improved
with further extensions. For example, we assume that all voters receive all
of their information through the media. While this is possible for people
who are unaffected by a natural disaster, it is unrealistic for those people
who are affected. The affected people experience the damage first-hand. A
possible simple extension would be to include two groups of voters: affected
who observe the disaster directly and unaffected who rely on the media.
This would be to follow the model in Besley and Burgess (2002). Our game,
however, concerns only the unaffected group, or those who learn about the
disaster through the media.
Also, we treat the loss from reporting the incorrect news as exogenous.
A possible extension would be to endogenize the loss, as in Besley and Prat
(2006). In their paper, t is a function of the number of other media outlets
that must be bribed, the media’s outside option from reporting, and trans-
actions costs between the politician and the media. This simple extension
would fit easily into our model.
Finally, our system of voting is not ideal. The voters vote based on what
the politician did last period. Ideally, voters would decide based on expec-
tations about next period. A simple extension would turn our model into
18
an adverse selection model. Different types of politicians will take different
actions, but the voters only know about the outcomes.
These simple extensions would provide a more realistic exposition of the
situation between politicians, media, and voters around a natural disaster.
19
Appendix
Finding equilibrium conditions, voter.
The voter plays a mixed strategy between voting for the incumbent and
voting for the challenger at each information set.
Let y be the probability that the voter selects the incumbent where he
receives a report of a disaster.
Let z be the probability that the voter selects the incumbent where he
does not receive a report of a disaster.
At ∅: Where the voter does not receive a report of a natural diaster, his
expected utility from voting in the incumbent is:
EU∅I = (1− πe − πt)(1− d)(b) + πe(1− s)(g) + πt(b−D)
and from the challenger:
EU∅C(1− πe − πt)(1− d)(x) + πe(1− s)(x) + πt(x−D)
Equating these expected utilities yields:
(1− πe − πt)(1− d)(b− x) + πe(1− s)(g − x) + πt(b− x) = 0
(V equilibrium condition 1)
At θ: Where the voter receives a report of a natural disaster, his expected
utility from voting in the incumbent is:
20
EU θI = (1− πe− πt)d(b−D) + πes(g −D)
and from the challenger:
EU θC = (1− πe − πt)d(x−D) + πes(x−D)
This yields the equality condition:
(1− πe − πt)d(b− x) + πes(g − x) = 0 (V equilibrium condition 2)
Equating these two conditions yields:
(1− πe − πt)(b− x) + πe(g − x) + πt(b− x) = 0
Rearranging
(b− x) = b− πeg
Solving for πe
πe =b− xb− g
Substituting πe into the condition at ∅ solves for πt
πt =(d− s)(g − x)
d(g − b)
21
Finding equilibrium conditions, politician.
The politician plays a mixed strategy, equating the expected utility from
each of each three actions, Ap ∈ N,E,R.
EUPN = dy(R) + (1− d)z(R) (Expected utility from playing N)
EUPE = sy(R) + (1− s)z(R)− e (Expected utility from playing E)
EUPT = z(R)− t (Expected utility from playing T )
Equating EUPN and EUP
E yields:
dy(R) + (1− d)z(R) = sy(R) + (1− s)z(R)− e
dyR + (1− d)zR = Rsy + rZ − szR− e
dyR− syR = dzR− szR− e (P equilibrium condition 1)
yR(d− s) = zR(d− s)− e
y = z − e
R(d− s)
Equating EUPN and EUP
T yields:
22
dyR + (1− d)zR = zR− t
dR(z − y) = t (P equilibrium condition 2)
y = z − t
dR
Equating EUPE and EUP
T yields:
sy(R) + (1− s)z(R)− e = z(R)− t
syR = e− t+ szR (P equilibrium condition 3)
y = z +e− tsR
Substiting the result from P equilibrium condition 1 into that for P equi-
librium 2, we have:
e
t=d− sd
23
N
N
M
M
M
M
V
V
V
V
V
V
V
V
Nothing
Effort Disaster^
Report*
Report*
Report*
Report*
Incumbent**
Incumbent**
Incumbent**
Incumbent**
Incumbent
P
Transfer
Incumbent**
Incumbent
Incumbent
R , l, b-D
0 , l, x-D
R-e , 0, g -e , 0, x
R-e , l, g
-e , l, x
R, 0, b-D
0, 0, x-D
R , 0, b
0, 0, x
0, l, x
R , l, b
R-e , l, g-D -e , l, x-D
-e , 0, x-D
R-e , 0, g-D
Extensive form game
ConAnued on the following page
22,23
24
N
M
N
VReport*
Incumbent**
M
V
VIncumbent**
M
V
Report*
Incumbent**
Incumbent**
M
V
V
Report*
Incumbent**
Incumbent**
Incumbent**
M
V
V
Transfer
Incumbent**
R, l, x-D 0 , l, b-D
R-t , t, b-D
-t , t, x-D
R , 0, x
0 , 0, b
R-t , t, x
-t , t, b
R , l, b-D
0 , l, x-D
R , 0, b-D 0 , 0, x-D
R , 0, b
0 , 0, x
R , l, b 0 , l, x
^ the complementary acAon is no disaster *complementary acAon is no report ** complementary acAon is voAng for a challenger
P: PoliAcian M: Media N: Nature V: Voter
22,23
25
References
Achen, C. and L. Bartels (2004). Blind retrospection: Why shark attacks
are bad for democracy.
Besley, T. and R. Burgess (2002, November). The Political Economy Of
Government Responsiveness: Theory And Evidence From India. The
Quarterly Journal of Economics 117 (4), 1415–1451.
Besley, T. and A. Prat (2006, June). Handcuffs for the Grabbing Hand?
Media Capture and Government Accountability. American Economic
Review 96 (3), 720–736.
Bovt, G. (2010). Putin’s vertical power disaster. The Moscow Times .
Cavallo, E. and I. Noy (2009). The economics of natural disasters: a survey.
Choe, C. and P. A. Raschky (2011). Media, democracy, and government
action: Prevention vs. palliation in the time of cholera. Iser discussion
paper, Institute of Social and Economic Research, Osaka University.
Eisensee, T. and D. Stromberg (2007). News droughts, news floods, and us
disaster relief. The Quarterly Journal of Economics 122 (2), 693–728.
Gasper, J. T. and A. Reeves (2011). Make it rain? retrospection and
the attentive electorate in the context of natural disasters. American
Journal of Political Science 55 (2), 340–355.
Guha-Sapir, D. et al. (2011). Annual disaster statistical review 2010. Cen-
tre for Research on the Epidemiology of Disasters .
John J. Tkacik, J. (2008, June). Seismic suppression: Chinese censorship
after the sichuan earthquake @ONLINE.
26
Kahn, M. E. (2005). The death toll from natural disasters: the role of
income, geography, and institutions. Review of Economics and Statis-
tics 87 (2), 271–284.
Leeson, P. T. and C. J. Coyne (2005). Manipulating the media. Institutions
and Economic Development 1 (2), 67–92.
Malhotra, N. and A. G. Kuo (2009). Emotions as moderators of infor-
mation cue use citizen attitudes toward hurricane katrina. American
politics research 37 (2), 301–326.
McMillan, J. and P. Zoido (2004). How to subvert democracy: Montesinos
in peru. Journal of Economic Perspectives , 69–92.
Porfiriev, B. (2014). Evaluation of human losses from disasters: The case
of the 2010 heat waves and forest fires in russia. International Journal
of Disaster Risk Reduction 7 (0), 91 – 99.
Quiroz-Flores, A. and A. Smith (2012). Leader survival and natural disas-
ters. British Journal of Political Science 43 (04), 821–843.
Sen, A. (1981). Poverty and famines: an essay on entitlement and depri-
vation. Oxford University Press.
Shvidenko, A. and D. Schepaschenko (2013). Climate change and wildfires
in russia. Contemporary Problems of Ecology 6 (7), 683–692.
Stromberg, D. (2007). Natural disasters, economic development, and hu-
manitarian aid. The Journal of Economic Perspectives , 199–222.
WashingtonBlog (2013, November). Japan reacts to fukushima crisis by
banning journalism @ONLINE.
27