Geoscience Reference
In-Depth Information
In this chapter, we focus on the household-level reaction to flood warnings in
Pakistan. Do early warnings change household behavior? Do different sources of
warnings prompt different types of mitigation behavior? Finally, do individual char-
acteristics, such as gender, impact individual abilities to take mitigation actions?
Our data allows us to reconstruct the 2010 Pakistan context of the flood warnings at
the village, household, and individual levels.
13.4
Empirical Methodology
We are interested in the effects of early warnings on household decisions to take risk
mitigation actions before the flood. To test the effect of early warnings on the likeli-
hood of taking mitigation action, we estimate the following equation using a logit
regression:
=+
(
)
+
(
)
M
αβ
warning
β
preparationtime
1
2
(
)
+
(
)
+ε
i
+
β
propensity
Σ β
other hhchars
(13.1)
3
i
M is an indicator variable that equals 1 if the household took mitigation action
(placing sandbag barriers around the property, moving possessions to higher ground,
reinforcing the house structure, or taking any of these three actions). Explanatory
variables include an indicator variable for whether a household had a warning, the
hours from the warning time until the flood waters entered the property, and the vil-
lage propensity for flooding that was calculated for village sample selection.
Controls include the age and gender of the household head respondent. Standard
errors are heteroskedasticity-robust and clustered at the village level.
To further examine the impact of such mitigation behavior on losses, we estimate
Eq.
13.2
using a Tobit regression, where the dependent variable L is the value of
losses.Lossvaluesarecensoredatalower-boundofzeroandanupper-boundof
totalpropertyvalue.Thismeansthatobservedlossescanneverbelessthanzero,
even if mitigation is infinite, and can never be higher than property value, even if
mitigation behaviors were to increase losses.
L=+
(
)
+
(
)
αβ
warning itigationaction
warningmitigationa
β
1
2
(
)
+
β
*
ction
3
(
)
+
(
)
+
+
β
propensity
Σ β
i
i
other
hhchars
ε
(13.2)
3
The warning coefficient measures the direct relationship of receiving a warning
with loss value, but we suspect that individuals who receive warnings are also more
likely to be those with high exposures and therefore higher losses. The interaction
term of warning and mitigation measures the marginal impact of receiving a warning
and taking any mitigation behavior.
Search WWH ::
Custom Search