Geoscience Reference
In-Depth Information
Finally, we estimate the effect of mitigation behavior on resilience, by estimating
Eq.
13.1
with a logit regression, where the dependent variable is an indicator vari-
able that equals 1 if the household has replaced losses by the time of the survey, and
the explanatory variables include mitigation actions.
All data was collected through a household survey of 640 households in Punjab
province in April 2013. The Punjab province is an advantageous location for sam-
pling both flood-affected and unaffected households, due to the five rivers flowing
through the province and the geographic diversity of flood effects. There was con-
siderable variation across the province in terms of rainfall levels, losses, and exter-
nal assistance.
Punjab province is divided into 36 districts, which are further subdivided into
127 tehsils.
2
Tehsils generally correspond to towns, but within one tehsil, there may
be multiple towns. Each tehsil is further divided into Union Councils that serve as
the local administrative units and can comprise multiple villages. The general meth-
odology followed by national surveys for rural areas is to then divide these villages
further into compact enumerator blocks of 200-250 proximate households out of
which 16 households are randomly selected for the survey. Using the terminology
from national surveys, we refer to these groups of 16 randomly selected households
as “clusters.” Our sampling frame is taken from a representative survey of 30,000
households in Punjab province that took place in 2011, about a year after the flood
had occurred. We sample flood-affected (“treated”) households and unaffected
(“control”) households that share similar characteristics along other dimensions that
could affect the outcome variables.
We expect to find variation in flood impacts, both between tehsils and within
tehsils, so our sampling strategy targets areas where we will be likely to find both
flood-affected households and adjacent unaffected households. We test not only the
direct impact of flood losses on risk perceptions and risk-taking behavior but also
the indirect impact of observing the flood event even without incurring personal
losses.Wethereforeselectdistrictswithvariationinlow/zero,moderate,andsevere
flood effects within the district.
To select districts that will allow sufficient variation of flood affected and non-
affected villages, we used the list of villages that were surveyed under the Multi-
cluster Rapid Assessment Mechanism (McRam) surveys 2010 and information
from the Multiple Indicator Cluster Survey (MICS) 2011 implemented by the
Punjab Bureau of Statistics (PBOS). The McRam, conducted in late August 2010,
was in 8 out of the 11 flood affected districts,
3
gathered detailed information on
flood damages and rehabilitation needs. In Pakistan, the MICS survey is imple-
mented approximately every four years in Punjab province. The survey draws a
sample of households from the total Punjab province population, representative at
3
According to the MICS 2011, the districts where any households reported being affected by the
loodsin2010wereRajanpur,Muzaffargarh,Jhang,Layyah,DGKhan,Sargodha,Multan,Rahim
Yar Khan, Bhakkar, and Bahwalpur.
Search WWH ::
Custom Search