Geoscience Reference
In-Depth Information
Next, we divide the list of matched villages into flood-affected and non-flood-
affected villages. Among the list of flood-affected villages, we randomly select
eightvillagesasthetreatmentgroup.WeselectfourvillagesinMuzaffargarh,two
in Layyah, and two in Rahim Yar Khan. Using the propensity scores, we map the
flooded villages and unaffected villages. For half (4) of the flood-affected villages,
we select the control village with a matching propensity score which is located at
closest proximity. For the other half (4), we select the control village with a match-
ing propensity score that is located at farthest proximity to flooding.7
7
For the “non-
flooded” villages, an additional check is performed using our several mapping
sources to verify that the village area was not considered flooded during 2010.
Five non-flooded villages adjacent to the flooded villages are selected from
Muzaffargarh,twoinLayyahandoneinRahimYarKhan.
We use the listing compiled for the latest round of the MICS 2011. It has the
complete listing for one randomly selected village block (blocks are a settlement
(
basti
) or a geographically concentrated group of households). We survey 20 house-
holds from each village.
To control for potential attrition bias of post-flood migration or absences, enu-
merators recorded reasons for why any household was not available for the survey.
Enumerators were provided with a list of five additional randomly selected house-
holds to draw for replacement, which was used when (i) no one was available who
can provide household information, (ii) the house structure was uninhabited, or (iii)
household members declined to participate in the survey. Given that the 2010 floods
induced temporary out-migration, there is a possibility that a sample of flood-
affected villages would under-represent flood-affected households. We use multiple
sources to approximate village population changes, and also ask about migration
directly in a separate survey module of a village leader who can give village-level
estimations. To directly collect information about village migration, we approach
each settlement with the pre-existing MICS roster of households, and we consult
with the village leader (“numberdar”) about the reasons why any households in the
listing may be missing now, and specifically the number of households that have
moved away from the area since 2010 due to the flood. In the household survey, we
also asked participants about the extent of migration in the village for various rea-
sons, including flood.
Table
13.1
shows summary statistics for the key variables of interest in the
household and individual surveys.
13.5
Results
Contrary to the perception that households did not heed warnings, we find that receiv-
ing any warning significantly increases the probability of taking mitigation actions.
However,weindthatcertainwarningtypesappeartobemoreeffectivethanothers.
7
The propensity scores of the non-flooded villages do not exceed the propensity scores of the
flooded villages by more than 30 % of the standard deviation of the scores.
Search WWH ::
Custom Search