Journal of Environmental Management 270 (2020) 110809

Contents lists available at ScienceDirect

Journal of Environmental Management

journal homepage: http://www.elsevier.com/locate/jenvman

Research article

Mitigating the impacts of floods using adaptive and resilient coping

strategies: The role of the emergency Livelihood Empowerment Against
Poverty program (LEAP) in Ghana
Martin Paul Jr. Tabe-Ojong a, *, Joyce Anima Boakye b, Mashauri Muliro c
a
Institute for Food and Resource Economics, University of Bonn, Nussallee 21, 53115, Bonn, Germany
b
Hochschule Bonn-Rhein-Sieg, Grantham-Allee 20, 53757, Sankt Augustin, Germany
c
World Bank, Kinshasa, Demoractic Republic of Congo

A R T I C L E I N F O A B S T R A C T

Keywords: Understanding how households make use of adaptive and resilient coping strategies are steps in the right di­
Floods rection of dealing with disaster risks and extreme weather and climate events. In the recent past, safety nets
LEAP programs have been quickly scaled up to respond to some of these climate-related and other types of shocks in a
Protection motivation theory
bid to improve resilience and restore the productive capacity of households. In Ghana for instance, there has been
Binary and count models
Flood resilience
the flagship Livelihood Empowerment Against Poverty (LEAP) program from which stemmed an ad-hoc cash
grant program, Emergency LEAP which is rolled out to victims of natural disasters like floods, storms, and
droughts. Based on the above, this paper analyzes the role of the Emergency LEAP in building the adaptive and
resilient nature of households against floods. With insights from an extended version of the Protection Motivation
Theory (PMT), the paper hypothesizes that relief assistance has a negative impact on the adoption of risk
mitigation measures against floods. The study employs both binary and count data models like the Probit, the
Poisson, and the Negative Binomial models respectively to establish the impact of such relief assistance on the
adaptive behavior of households.The findings reject the null hypothesis; thus, strengthening the argument that
relief assistance has a negative influence on the use of resilient coping strategies. Furthermore, the study es­
tablishes that experience in floods increases the resilient nature of households. The study concludes that social
safety net programs like the emergency LEAP could perform better if given a renewed focus that transcends
beyond just cash benefit to including much better long-term interventions like the integration of community-
driven development (CDD) programs as well as inducing behavioral change through information access and
education.

1. Introduction National Disaster Management Organization (NADMO) which is

responsible for numerous tasks such as the organization and coordina­
Climate change and disaster risks are major concerns globally and tion of relief assistance to victims (NADMO, 2016) and the Environ­
countries, governments, communities, and individuals all over the world mental Protection Agency (EPA), responsible for improving and
are in one way or the other commiting and resorting to measures to deal protecting the environment (air, land, and water) in Ghana (EPA, 2018).
with their looming effects (Barros et al., 2014; Kuwornu, 2019; UNEP Another example is the Ministry of Environment, Science, Technology,
&UNDP, 2012). This commitment has seen countries like Ghana signing and Innovation which oversees the development and promotion of sci­
agreements under the United Nations Framework Convention on ence and technology to ensure safe and sound environmental practices.
Climate Change (UNFCCC), including ratifying the Paris Climate Accord It is very evident that disaster impacts are felt by everyone, however,
and the Kyoto Protocol as well as adapting national laws and policies vulnerable and poor households living in underprivileged conditions are
such as the National Climate Change Adaptation Strategy (NCCAS). In more affected based on their dependency on agriculture and its related
essence, these laws mandate specific ministries and agencies to deal with activities for their livelihood. Factually, climate change and extreme
both climate change and disaster risks. Amongst these agencies are the weather events are pushing people into poverty (Birkmann et al., 2013;

* Corresponding author.
E-mail addresses: [email protected], [email protected] (M.P.Jr. Tabe-Ojong).

https://doi.org/10.1016/j.jenvman.2020.110809
Received 3 June 2019; Received in revised form 21 December 2019; Accepted 19 May 2020
Available online 8 July 2020
0301-4797/© 2020 Elsevier Ltd. All rights reserved.
M.P.Jr. Tabe-Ojong et al. Journal of Environmental Management 270 (2020) 110809

Hallegatte et al., 2018; World Bank, 2013). For this reason, it is in building households’ adaptive behavior towards floods. The analysis
imperative to harness efforts towards building synergies between three can be seen as a partial test of this theory with data from Ghana. As a
vital sectors namely, social protection, climate change adaptation (CCA) matter of fact, studies linking PMT to natural hazards like floods are rare
and disaster risk reduction (DRR) to help households build better resil­ in the context of the global South (Grothmann and Reusswig, 2006).
ience against risks that are predominantly triggered by climate change This study thus offers fresh insights while triangulating existing facts
and disasters. When these three sectors are harnessed well, there will be and hypotheses from the global North.
efficient and effective delivery of relief assistance to victims while Secondly, it builds on the theoretical contribution of the extant
improving the knowledge base and information access of households. literature by applying both binary and robust count data models like the
Also, there will be the provision of the required technical and capacity Poisson and the Negative Binomial model to understand what drives
building support necessary in preventing households from falling back households to use resilient protective measures against floods and the
into such shocks. intensity of the measures used. Most previous analyses end at under­
Mitigating the negative impacts of extreme weather events such as standing what drives households to use any of these measures (Mabuku
floods with adaptive and hands-on coping strategies are pro-active et al., 2018; Poussin et al., 2014; Richert et al., 2017) without extending
measures to improve the livelihood of households. While the role of to the level (intensity) of use of these coping measures. Households
both integrated and comprehensive flood risk management measures usually engage in a couple of mitigating measures like purchasing in­
are well established in extant literature (Grothmann and Reusswig, surance, building brick walls and fences, purchasing of emergency
2006; Mabuku et al., 2018; Osberghaus, 2017; Poussin et al., 2014), health kits with some households even permanently relocating
there still exists a dearth of information on what triggers households to (migrating). Understanding what drives households in using a number
use these coping and adaptive measures against environmental shocks of these measures is very crucial and relevant for policy evaluation and
like floods. Empirical analyses thus far have largely focused on the role development.
of human capital characteristics like experience and education (Osber­ Although the paper focuses on understanding the coping and adap­
ghaus, 2017) in mitigating these shocks through the use of various tive behavior of households, it does not engage fully into the technical
coping and adaptive strategies. and physical aspects of resilience, nor does it attempt to prioritize spe­
Osberghaus (2017) examined how households augment their miti­ cific risk mitigating measures.
gation efforts against floods after they have experienced a flood threat in
Germany. Employing the difference in difference estimator, the study 3. Emergency LEAP and disaster risk management in Ghana
established a causal relationship between flood experience and the
mitigation strategies employed by households. Education was also found Ghana, like all other developing countries, is challenged by the harsh
to play a crucial role in building household resilience: the higher the effects of climate change and extreme weather events. Climate effects
level of education of the head of the household, the more responsive range from an increase in average annual temperature of approximately
they are to floods. Contradictory insights were earlier provided by 1� which represents an average increase of 0.21 � C per decade and an
Poussin et al. (2014) who rather found non-flood experienced house­ increase in the average number of hot nights per year (MESTI, 2013;
holds using more mitigating measures than flood experienced house­ Herrera and Flora, 2017). The high temperatures have over the years
holds. They argued that households employing mitigating strategies caused the rise in sea levels causing floods and coastal erosion of 1.13m
may probably have little or no experiences with the downsides of floods per year in major coastal areas in the country (MESTI, 2013; Herrera and
as they already implemented the necessary actions. Flora, 2017). Other recorded disasters are excessive drought which has
Against this background, this paper seeks to provide insights into this impacted the production of hydropower supply for over six years
ambiguity, exploring the relationship between Emergency LEAP and (Australian AID, 2012). It has also impacted agriculture with sometimes
flood mitigating measures. It is structured as follows: section two limited or no rain for farming which threatens the income and livelihood
highlights the aims and scope of the paper while section three offers a of farmers (Australian AID, 2012).
brief insight into Emergency LEAP and disaster risk management in the Disaster management in Ghana is carried out by the National
context of Ghana. Section four provides the theories and concepts Disaster Management Organization (NADMO) with the support of na­
starting with the extended PMT and going to the model specification. tional and international bodies (NADMO, 2016). NADMO was estab­
Sampling, study site and data collection issues are discussed in section lished as a result of the United Nations Office for Disaster Risk
five. The article proceeds with both descriptive and econometric results Reduction-UNDRR (formerly the International Strategy for Disaster
in section six and ends with the conclusion in section seven. Reduction or ISDR) conference held in Yokohama, Japan in 1994 to
draft a disaster risk reduction strategy (NADMO, 2016). It performs
2. Aims and scope several tasks and key amongst them is disaster prevention, risk, and
vulnerability reduction as a means of reducing the societal impact of
Taking into account the scientific debate on the relationship between disasters (NADMO, 2016). In line with the Hyogo Framework for Action
flood experience, education and the flood mitigating strategies of 2005–2015, NADMO developed a comprehensive plan for the preven­
households, this study seeks to establish the evidence linking financial tion and management of disaster and climate associated risks called
incentives in the form of social payments and the adaptive behavior of “The Ghana Plan of Action on DRR and CCA” (NADMO, 2016). This plan
households against floods. This is the gap this paper tries to fill by has NADMO as the coordinating and implementing body and collabo­
examining the role of the Emergency LEAP in building the mitigative rates with many stakeholders for its execution. One of the many stake­
behavior of households in Ghana. Specifically, the study analyzes how holders is the then Ministry of Women and Children Affairs (MOWAC),
EmergencyLEAP payments impact the use of resilient mitigating mea­ currently the Ministry of Gender, Children and Social Protection
sures. It also tests other factors that might have an influence in building (MoGCSP) (NADMO, 2016). According to the plan, MOWAC amongst
the coping and adaptive strategies of households against floods. other things is responsible for women sensitization through disaster risk
The study employs the extended Protection Motivation Theory reduction and climate change adaptation training as well as resourcing
(PMT) to understand the severity of the flood (threat appraisal) and how them in viable activities to build assets and improve on their coping
households build resilient measures against these floods (coping mechanisms.
appraisal). Specific attention is given to the coping appraisal since it With the name change in the ministry, there is a renewed focus on its
underscores the efficacy and self-efficacy of households in building their aim which includes more comprehensive programs geared towards
coping strategies. The contribution of this study is thus twofold: firstly, it poverty reduction. As part of Ghana’s social protection mandate to
offers primer and empirical insights into the role of financial incentives protect all vulnerable and disadvantaged households, MoGCSP

2
M.P.Jr. Tabe-Ojong et al. Journal of Environmental Management 270 (2020) 110809

introduced a temporary cash transfer program aimed at poverty eradi­ appraisal. While the threat appraisal profiles and examines the severity
cation with a focus on reducing disaster-related poverty known as of an event, the coping appraisal refers to how individuals respond or
Emergency Livelihood Empowerment Against Poverty (Emergency better still adapt to a situation. The coping appraisal is thus subdivided
LEAP) program. As NADMO oversees the evacuation services and pro­ into efficacy and self-efficacy. Efficacy is the expectation an individual
vision of in-kind support, the Emergency LEAP renders cash support to has that adherence to specified recommendations can remove or reduce
victims. It has created a disaster response mechanism that transfers the effects of a threat. Self-efficacy, on the other hand, is the belief in
money to disaster victims across the country through an established one’s strengths and self to follow the specified recommendations. The
payment system. The current focus of Ghana’s social protection on coping appraisal has been found to have a greater influence on mitiga­
reducing disaster-related poverty is based on global and national shred tion behavior than the threat appraisal (Poussin et al., 2014). Overall,
of evidence that have proven that disasters exacerbate poverty and for PMT is based on the idea that individuals will protect themselves against
that matter, make it difficult to achieve stated national social protection risks if the perceived threat has a high severity and the efficacy and
objectives. self-efficacy of the individuals are also high. As part of the coping
As Ghana continues to face one disaster after the other, the govern­ appraisal, individuals have to be sure that the implementation of the
ment in collaboration with notable stakeholders like NADMO and the mitigating measure is not very costly. Otherwise, this may act as a
Social Protection ministry realized the importance of setting up a hindrance and reduce the self-efficacy of the individual.
secured, safe and efficient means of getting cash support to victims. For PMT was originally formulated to profile and better understand how
this reason, there was the need for the creation of a structured national individuals adapt and cushion themselves against health risks (Rogers,
cash transfer program called Emergency LEAP which is under the aus­ 1975). It has recently been further developed in the context of flood risks
pices of Ghana’s flagship cash transfer program, LEAP. Emergency LEAP (Grothmann and Reusswig, 2006). Fig. 1 demonstrates the extended
is an ad hoc cash transfer program established in 2010 as a result of the version of the PMT beginning from the basic threat and coping appraisal.
influx of severe drought and floods in many districts across eight out of In the extended framework, issues of risk and flood experience, social
the ten regions of Ghana. It is tasked with addressing the temporary networks, and social norms, knowledge and information characteristics,
basic needs of disaster victims across the country and restoring their as well as socio-economic characteristics, are brought into scenario as
productive capacity. In this program, people affected by floods are ex­ they are thought to influence and build the adaptive and risk mitigating
pected to receive cash payments up to a period of three cycles and then behavior of individuals. Individuals who have previously incurred flood
enroll them in the LEAP program shoud they be non-beneficiaries. As of damages and lost some properties are expected to invest more in their
2015, a total of 11,102 households from five regions had benefited from use of mitigating strategies against floods. In the same vein, reliance on
the Emergency LEAP: Greater Accra 6,996, Volta 996, Ashanti 842, public flood protection programs like social payments can negatively
Central 1307 and Western 961. affect the use of such precautionary measures (Richert et al., 2017). This
is because flood mitigation can be seen as a behavioural concept where
4. Theories and concepts the perceptions, emotions and effort level of households matter
(Osberghaus, 2017). Thus any external individual support which is not
4.1. Protection motivation theory geared at changing the behavior of individuals will only enhance a
negative impact of the individual payments on the use of risk mitigating
The conceptual framework is based on the extended version of the measures.
Protection Motivation Theory. The protection motivation theory (PMT) Depending on the kind of hazard or shock, individuals respond
hinges on the fact that households protect themselves from shocks and differently in a bid to reduce the negative impacts of the event. There are
risks based on four distinct factors: the perceived probability that an basically two response types, protective responses and non-protective
(unlikely) event will occur, the severity of the event, the efficacy of responses (Grothmann and Reusswig, 2006). Protective responses are
coping/adaptive behavior and the self-efficacy of the households. It is geared towards reducing or preventing losses when an event occurs.
thus a combination of both the threat appraisal and the coping/adaptive Non-protective response, on the other hand, involves some sort of

Fig. 1. Protection Motivation Theory (Extended version) (Source: Poussin et al. (2014)).

3
M.P.Jr. Tabe-Ojong et al. Journal of Environmental Management 270 (2020) 110809

fatalism. Thus, it is more psychological and includes attitudinal char­ estimates for the regression coefficients as it assumes the zeros jointly
acteristics like fear. This is usually the case if the individual’s coping emerge from both a binomial process and a count process.
appraisal is low especially when monetary costs are very high. Choosing In our case, zeros represent the decision of households to be fatalistic
a protective response does not imply action especially in the presence of and not invest in any risk-mitigating measures, owing to a plethora of
resource barriers like time, income and knowledge. reasons. In such cases, what comes to mind is the Poisson model.
However, it assumes that the variance and mean of the distribution is
equal, which is usually not the case in practice. The Poisson model is
4.2. Model specification represented as:

In every empirical analysis, the choice of an appropriate estimation e u uy

pðY ¼ yÞ ¼ (4)
model is usually determined by the nature of the dependent variable. y!
Our dependent variable is a binary variable that takes the value of 1 if a
where uis the intensity or rate parameter and delineated as
household used any flood protective measure and 0, otherwise. Logit
�
and probit models are the most commonly used models for binary u ¼ exp x’i δ (5)
dependent variables. Since the models yield similar results, we can es­
timate with any of them. To understand the drivers of the use of a risk- The Poisson model follows the Poisson distribution which is based on
mitigating measure, we employ the probit model as shown below. the fairly restrictive equidispersion assumption (Pittman et al., 2018).
Assuming the equality of the mean and the variance, the equidispersion
Y* ¼ Ф 1ðYÞ ¼ x’i δ ​ þ ​ ε (1) property can be expressed as:

where Y * is a latent variable representing the utility differences in the use EðyjxÞ ¼ varðyjxÞ ¼ u (6)
of a resilient risk-mitigating measure, such that; Because of its restrictive properties, the Poisson model is less
� commonly used in empirical analysis as overdispersion is very common
1 if Y * � 1
Y¼ (2) in most datasets. The Negative Binomial (NB) is a more flexible alter­
0 otherwise
native that discards this equal mean and variance assumption, allowing
x’i is a vector of explanatory variables thought to influence the use of the variance to be greater than the mean. It does this by introducing a
resilient risk protective measures, Ф is the cumulative normal distribu­ dispersion parameter, α
tion wherein ФðzÞεð0; 1Þ, δ is a vector of parameter estimates and Ɛ is the varðyjxÞ ¼ u þ αu2 (7)
stochastic error term which follows a normal distribution, Ɛ � Nð0; 1Þ
The cumulative density function of the normal distribution for the The NB model estimates the α parameter and also tests if it is sta­
probit model can be represented as: tistically significant from zero through a likelihood ratio test. If α ¼ 0,
the NB model converges to a Poisson model, implying the Poisson model
Z
is a nested NB model.
� � x’i δ
F x’i δ ¼ Ф x’i δ ¼ φðzÞdz (3)
∞ For brevity of interpretation and purposes of comparison, we will
derive the marginal effect of both the Poisson and NB model which is
While we recognize the ability of the probit model to estimate
expressed as:
whether households use resilient protection measures or not, it is
important to mention that households usually undertake more than one ƏEðyjxÞ �
¼ δj exp x’i δ (8)
risk mitigating measure. Limiting the analysis to a binary decision of Əxj
households may lead to incorrect and misguided inferences. Moreover,
as different households use a combination of different risk mitigating The choice of explanatory variable to be used in the model is based
measures, it becomes worthwhile to understand what drives their de­ on theoretical and extant empirical analysis. Some of these variables
cision in doing so. Thus, we move above the basic binary decision to include household socioeconomic and biographic variables such as ed­
understand what influences the intensity (in our case the number) of ucation, age experience as well as the payment of Emergency-LEAP and
mitigating measures used. Our dependent variable can now be seen as a the level of destruction of the previous flooding shock.
count variable of the number of risk mitigating measures used by a
household. This implies different count variable models like the Poisson, 5. Study site and data collection
zero-inflated Poisson, negative binomial model, zero-inflated negative
binomial and the hurdle Poisson model (Pittman et al., 2018) can be The study population consists of flood-prone areas in the Volta re­
used to estimate what influences households to use different measures to gion of Ghana. The Volta region of Ghana is a typical area for flood
enhance their resilience to floods. Using any of these models is based on hazards in the country. However, the tidal wave we are considering in
its ability to cater for zeros in the data. The model should also be able to the analysis occurred in the Ketu South district. The once commercially
control for overdispersion or underdispersion where the variance may vibrant district is geographically located at the eastern end of a narrow
be larger or smaller than expected with some skewness. strip stretching from the Volta estuary to the keta lagoon complex,
Modeling the source of zeros in any model is very important as zeros which is a designated Ramsar site (WACA, 2017). The land surface area
usually emerge from different sources and processes. In some cases, of the municipality is 30% covered by watery bodies including the Keta
zeros may emerge from the decision of the household, in which case it is Lagoon which is 12 km wide and 32 km long. The study was conducted
a corner solution to the household’s utility maximization problem. In specifically in Blekusu and Agavedzi in the Ketu South district (Fig. 2).
such cases, a Tobit model or the more flexible hurdle models would be The district covers a total land area of 278 km2 and according to the
more appropriate. In other cases, zeros may arise as a result of incidental 2010 population and housing census, the district has a total population
truncation where a part of the variable is unobserved due to sampling of 160,756 with 75,648 males and 85,108 females (GSS, 2013; Thomas
techniques and sampling issues. This is a form of sampling selection bias Brinkhoff, 2018). Agavedzi and Blekusu are predominantly a rural
and sample correction models like the Heckman’s two-stage model or fishing area with a combined total population of about 36,409 in­
better still the Heckman’s correction caters for these zeros. In other habitants. In 2017, residents of these towns were displaced by strong
cases, the zeros may come from different data generating processes, in tidal waves which according to the residents are an annual ritual. To
which case we refer to them as being inflated zeros or with extra zeros. reduce the impacts on the affected households, the National Disaster
Models like the zero-inflated Poisson provide consistent and unbiased Management Organization assisted in getting temporal shelters for the

4
M.P.Jr. Tabe-Ojong et al. Journal of Environmental Management 270 (2020) 110809

Fig. 2. Map of study area.

displaced and took stock of their losses. The Ministry of Gender, children
and social protection (MoGCSP) later made cash disbursement to some Table 1
Variable Description and a priori Expectation.
of the displaced households as part of its Emergency LEAP mandate.
A simple random sampling technique was employed where 200 Variable Measurement A priori
expectation
households were randomly selected both from Blekusu and Agavedzi.
Interviews were used to elicit responses from the households through Dependent variables
the use of a well-structured survey instrument. Our survey instrument Protective Dummy ¼ 1 if household used any resilient
Measure (PM) social protection measure, 0 otherwise
had questions ranging from socio-economic characteristics of house­ Number of PM Number of protective measures used
holds, flooding experience, and the properties lost and coping strategies Outcome variables
that the households used to whether or not they received the Emergency Payment Dummy ¼ 1 if the household received þ
LEAP payment. received payment, 0 otherwise
Gender Dummy ¼ 1 if household head is male,
Table 1 shows the variable description, their measurement units and þ/
0 otherwise
the expected signs for the explanatory variables in the model equations. Age Age of the household head in years –
Households who received the Emergency LEAP payment are expected to Marital status Married ¼ 1, single ¼ 2, Divorced ¼ 3, þ/
be more proactive and invest in risk mitigating measures than house­ Widowed ¼ 4
holds without the payment. Like subsidies, social payments are always Occupation Categorical ranging from Fisherman ¼ 1, þ
vocational jobs ¼ 2, job trader/fish monger
intended and targeted at improving the resilience of households against ¼ 3, student ¼ 4, educational job ¼ 5,
shocks. In this case, the payment was not only made to ensure house­ Others ¼ 6
holds emerge from the flood shock, but also to make them more resilient Educational level Number of years in education þ
against emerging flood shocks. The payment was planned to be carried Properties lost Number of properties lost by household due þ
to floods
out on a three-round rolling basis with a cash payment of 300 GH’1 each
Payment Number of times payment was done þ
round. Thus the frequency of payment is a priori expected to increase the frequency
investment of households. Assumably, households who received the Household Income of household in Ghana cedis þ
payment more than once will get more income for investment into flood income
protective measures compared to households who received the pay­ Blekesu Dummy ¼ 1 if household is in Blekesu þ
village, 0 otherwise
ments on a single basis or not at all.
Agavedzi Dummy ¼ 1 if household is in Agavedzi
The gender construct of households is also thought to have an impact village, 0 otherwise
on the use of flood protective measures. The sign of this impact is rather
ambiguous as adaptive measures are a behavioral concept which cannot
be effectively profiled under gender. Age, on the other hand, is expected acceptance of novel innovations with proven benefits.
to have a negative effect on both the use and number of adaptive mea­ Unlike the gender variable, the marital status of households is ex­
sures adopted by the household. While age represents a build of expe­ pected to depict a positive relationship with the use of risk-mitigating
rience and an improved knowledge base on various adaptive measures social protection measures (Mabuku et al., 2018) with married re­
as well as novel ways to protect households from impending shocks and spondents having a greater probability of using risk mitigation mea­
risks, younger households are rather expected to use these adaptive and sures. The livelihood activity of the household head is expected to also
novel risk-mitigating measures more than their older counterparts. This have an ambiguous influence on the use of these risk-mitigating mea­
behavior can be attributed to their less risk-averse nature and greater sures. Households who are into fishing and other vocational jobs are
expected to undertake similar if not the same coping and adaptive
measures with households who are into other educational jobs.
1
1 GH’. ¼ 0.19USD as at 03/06/2019. The educational level of the household head is expected to have a

5
M.P.Jr. Tabe-Ojong et al. Journal of Environmental Management 270 (2020) 110809

positive impact on the use of risk mitigating and adaptive measures given thrice was only given once with a general and fixed amount of 300
against impending flood shocks, since education increases the accumu­ GH’. The gender of households was considered as the gender of the
lation of knowledge and the processing time of relevant information household head. It takes on the value of 1 for male-headed households
(Atreya et al., 2017; Mabuku et al., 2018; Osberghaus, 2017). Infor­ and 0 otherwise. From the results, 71% of the households are male-
mation is very relevant for the adoption and use of risk-mitigating headed. Age is also considered as the number of years of the house­
measures. As a matter of fact, households always network and share hold head. The mean age of most household heads is approximately 49.5
experiences on the best measures and technologies to use (Atreya et al., years implying most of the household heads are in the middle-aged
2017). Moreover, education can be seen to reduce the risk aversion of group.
households as they consider the benefits vis-a �-vis the losses if action is The marital status of the household is also considered and treated as
not taken. The same is expected of the properties lost as a result of the a categorical variable taking the values of 1, 2, 3 and 4 for married,
previous floods. Households that lost properties are expected to invest in single, divorced and widowed couples respectively. From the descriptive
risk-mitigating measures as a means of preventing future losses (Richert results, a great majority of the households are married. Also, the prop­
et al., 2017). Households usually develop risk and shock mechanisms erties previously lost as a result of the floods are considered. Households
based on the risk they face and the losses they incur. As a result, reported losing items ranging from a complete house, part of the house,
households with more properties lost would be less fatalistic about the household equipment, food, and fence to their source of livelihood.
future and invest in more risk-mitigating measures. From the household income viewpoint, the mean household income
The income of the household is a priori expected to have a positive is 651.5 GH’, with most households securing their livelihood from
effect on the use of risk-mitigating measures (Grothmann and Reusswig, various economic activities including support and remittances from
2006). Using some of these adaptive and resilient coping measures are abroad.
usually very costly and requires the use of different income sources.
Higher income households will invest more in these measures than low
6.2. Probit model of the determinants of use of risk mitigating protection
income households, keeping everything constant.
measures
6. Results and discussion
Table 3 presents the results of the factors influencing the use of a risk
mitigating measure against floods. The likelihood ratio test is indicative
6.1. Descriptive results
of the fact that the slope coefficients are statistically significant from
zero for the decision to use a risk mitigating measure.
The variables employed in the study can be grouped into household
The coefficient of payment received is negative and significant at the
socioeconomic, wealth, institutional and asset variables. Table 2 offers
5% level of probability with the use of a risk mitigating measure.
summary statistical parameters such as mean, standard deviation and
Households who received the Emergency LEAP payment are less likely
range values of the variables considered in the empirical model. Our
to use flood mitigating measures than households who did not receive
dependent variable, the use of a risk-mitigating measure is both treated
the payment. This is highly expected and in line with our conceptual
as a binary variable and a count variable. As a binary variable, it takes
model which contended that public policy support, especially in the
the value of 0 or 1 with 1 representing households that use any risk-
form of payments, has a negative impact on the use of risk mitigation
mitigating measure and 0 otherwise. It also has a mean which is
measures. This is especially the case if these payments are directed to­
greater than 0.50 implying about 65% of households in the study area
wards households like in the case of Ghana. Moreover, households might
are using at least one risk-mitigating measure. As a count variable, it has
have invested this payment into other uses. Mindful of the fact that most
six different risk-mitigating options employed by households; ranging
of these households are agrarian with fishing being their main livelihood
from the construction of a brick wall around their settlements, pur­
activity, the benefit received may have probably been invested in fishing
chasing of insurance, purchasing of stronger building materials and
activities. As a matter of fact, the payment was meagre (300 GH’) and
enhancing their settlements, temporary relocation arrangements, com­
not given three times as was previously planned. Households may hav
munity action like digging furrows and finally a combination of resource
invested these payments in restoring their livelihoods.
availability and flood preparedness items. These include the availability
Meeting a priori expectation, the coefficient of properties lost during
of physical items like emergency kits, boats, and food.
previous flooding shocks is significant and depicts a positive relationship
The main variable, the benefit received is considered as a dummy
with the use of a protective measure. Households that were previously
that takes on the value of 1 for households that received the Emergency
struck with the floods and incurred some losses have a greater likelihood
LEAP payment and 0 otherwise. It has a mean value of 0.68 implying the
of using protective measures than households that recorded no losses.
proportion of households that received the payment is more than the
proportion that did not receive. The payment which was intended to be
Table 3
Determinants of the use of Risk Mitigating Protection Measures.
Table 2
Variable Coefficient Standard error
Model variables and summary statistics.
Payment received 0.167** 0.719
Variable Mean SD Min Max
Gender 0.019 0.080
Payment received 0.68 0.46 0 1 Properties lost 0.033* 0.002
Protective Measure (PM) 0.65 0.47 0 1 Educational level 0.005 0.008
Number of PM 2.43 1.98 0 6 Age 0.003* 0.002
Gender 0.71 0.45 0 1 Household income 0.0001 0.0001
Age 49.44 17.71 20 100 Agavedzi village 0.197 0.066
Marital status 1.54 0.87 1 4 constant 0.876* 0.506
Occupation 2.59 1.47 0 6 Pseudo R2 0.049
Educational level 4.32 4.42 0 16 Log-Likelihood 119.845
Properties lost 2.23 1.69 1 6 Prob(Chi2 ) 0.08
Payment frequency 0.70 0.46 0 1
LR Chi2 (7) 12.50
Household income 651.5 251.8 100 1000
Blekesu village dummy 0.49 0.50 0 1
Notes: ***significant at 5%, **significant at 10%. Source: Authors computation
Agavedzi village dummy 0.51 0.50 0 1
from field survey (2018).
Source: Authors’ computation from field survey (2018) Source: Authors computation from field survey (2018)

6
M.P.Jr. Tabe-Ojong et al. Journal of Environmental Management 270 (2020) 110809

Mathematically, a unit increase in the number of properties lost in­ Table 5

creases the probability of the household using resilient measures by Determinants of the intensity of use of flood protective measures-negative
0.03. This points to the fact that households are more reactive than binomial model.
proactive in their behavioural decisions. Variable Negative Binomial model Marginal Effect
The age coefficient is also found to be negative and significant at the Coefficient Standard Coefficient Standard
10% level of probability. Younger households take more preventive error error
protection decisions and are more risk preventive than their older
Payment received 0.286** 0.146 0.696* 0.362
counterparts. This is in line with previous studies that have found Gender 0.035 0.159 0.085 0.387
younger households to be more risk averse, especially in the face of Properties lost 0.083** 0.041 0.202** 0.103
novel technologies and adaptive mechanisms to various daunting envi­ Educational level 0.006 0.018 0.015 0.044
ronmental and food production issues. The constant term is also positive Age 0.008** 0.004 0.021** 0.010
Household 0.0001 0.0002 0.0002 0.0007
and statistically significant implying that keeping everything constant, income
households have a greater probability of using risk-mitigating measures Blekesu village 0.121 0.138 0.295 0.339
to minimize the impact of an impending risk or shock. constant 1.177*** 0.361
Pseudo R2 0.0143
Log-Likelihood 395.186
6.3. Factors influencing the intensity of risk mitigating measures used by Prob(Chi2 ) 0.0000
households LR Chi2 (7) 11.46
Alpha 0.493*** 0.130
The Poisson model which shows the determinants of the number of
Notes: *** significant at 1%, ** significant at 5%, * significant at 10% Source:
risk-mitigating measures used by a household is presented in Table 4.
Authors computation from field survey (2018).
For comparison purposes, the negative binomial is also presented in
Table 5. All the models are estimated and their marginal effects obtained
with the use of the margins command in STATA. The marginal effect implementation of risk mitigating and coping measures. This is even
depicts the effect of an incremental change in a particular covariate on more evident when considering the value of the constant term (1.177)
the number of measures adopted. The two models offer very similar which is also significant at the 1% level of probability. The constant term
results, especially in terms of the overall significance of coefficient es­ always depicts the value of the dependent variable (in our case, the
timates. However, the NB model offers the alpha parameter which is number of flood protective measures used) if all confounding factors are
highly significant at a 1% level of probability, implying that the NB kept constant. Thus, households are generally willing to invest in risk
model is a better fit to the data than the Poisson model. Moreover, the mitigating measures as a form of adaptation and mitigation against
log-likelihood ratio is also significant at the 1% level of probability with future shocks and risks.
a value of 11.46. Based on the better fit of the NB model, this study The coefficient of property lost is also significant (p < 0.05) and
extensively discusses insights based on the NB model. depicts a positive relationship with the number of flood protective
Like in the probit model, the payment received, properties lost and measures used. Households who hitherto lost their properties to floods
the age covariate are significant determinants of the number of flood have a greater likelihood of increasing the number of flood protective
protective measures used by a household. The main covariate, the measures used. The age of the household head is also negatively corre­
payment received is negative and significantly different from zero. This lated with the number of flood protective measures used like in the
is interpreted to mean that for any additional payment made to house­ probit model. Most of the flood protective measures used by the farmers
holds, the difference in the logs of expected counts would decrease by can be viewed as adaptive and highly novel. Households with younger
0.286, implying that households who receive the payments are less heads are therefore expected to use more of these protective measures
likely to adopt any of the risk mitigating measures than households who than households with older heads. The younger the household head, the
did not receive the payment. The marginal effect tells us that households more risk-averse they are which leads to a higher probability of
who received the payment are also likely to have their use of additional investing in novel environmentally resilient measures.
protective measures reduced by 69.6 percent. These results are in line
with the PMT framework presented by Richert et al. (2017) in that, 7. Conclusion
reliance on public flood protection negatively affects the
The impact of climate variability and climate change are increasingly
being felt as extreme weather events like floods keep rising. Under­
Table 4
standing the coping and adaptive behavior of communities exposed to
Determinants of the intensity of use of protective Measures-Poisson model.
flood risk is relevant for policy development and stakeholders bent on
Variable Poisson model Marginal Effect
building the resilience of communities to natural hazards like floods.
Coefficient Standard Coefficient Standard While it is well established in the literature that flood experience, edu­
error error cation and social network characteristics of a community are relevant in
Payment received 0.270*** 0.096 0.641*** 0.225 building their resilience to natural hazards and in particular floods, very
Gender 0.055 0.108 0.132 0.257 little is known on the role of financial incentives in building the resil­
Properties lost 0.075*** 0.025 0.178*** 0.060
ience of households against floods. We used a financial relief program,
Educational level 0.006 0.012 0.014 0.029
Age 0.008*** 0.003 0.020*** 0.007 the Emergency LEAP which was launched in 2010 in Ghana. It is a build-
Household 0.0006 0.0001 0.0001 0.0004 up of the LEAP but targets only disaster victims. While the impacts of the
income LEAP have been considerably carried out, very little is known of the role
Blekesu village 0.116 0.0923 0.296 0.218 of the Emergency LEAP in building the adaptive nature of households
Constant 1.198*** 0.245
0.0294
towards floods. We thus bridge this gap by providing empirical evidence
Pseudo R2
on the role of the Emergency LEAP in triggering the use of various
Log-Likelihood 412.464
Prob(Chi2 ) 0.0008 coping strategies against floods in Ghana.
LR Chi2 (7) 25.02 The study hypothesizes that financial payments have a negative
impact on building the adaptive and resilient nature of households
Notes: *** significant at 1%, ** significant at 5%, * significant at 10% Source: against floods. With insights from the protection motivation theory, the
Authors computation from field survey (2018).

7
M.P.Jr. Tabe-Ojong et al. Journal of Environmental Management 270 (2020) 110809

study tests this hypothesis using both binary and count data models to Investigation, Methodology, Writing - original draft, Writing - review &
establish what drives the decision of the household to use a risk miti­ editing. Joyce Anima Boakye: Conceptualization, Data curation,
gatingmeasure and the number of risk mitigating measures respectively. Investigation, Writing - original draft, Writing - review & editing.
The study rejects the null hypothesis; thus, strengthening the argu­ Mashauri Muliro: Writing - review & editing.
ment that relief assistance has a negative impact on the use of risk
mitigating coping strategies. This stems from the fact that risk mitigation Acknowledgements
and risk coping strategies are a behavioral concept wherein success can
be achieved faster by changing or altering the behavior of households. The authors are grateful to Brot für die Welt through the Evangeli­
Changing the behavior can be obtained through the role of education, sche Studierendengemeinde (ESG) Bonn for funding the data collection
information and building capacity. Financial incentives alone may not in Ghana and the LEAP secretariat for assisting in the sampling and data
yield the intended effects if there is no induced behavioral change. As a generating process. Special thanks to Ogunla Jeremiah for generating
matter of fact, the payment which was planned to be given three times the map of the study area. The views expressed in this publication are
was only provided once. While this may broadly indicate the limitations those of the author(s) and do not necessarily reflect the views or policies
of the program in fully meeting its objectives in terms of full payments, it of the organizations they represent.
may also indicate unmet promises which to a large extent significantly
reduces the intended benefits of the program. In a way, our findings References
support this premise as most households used the payments for other
purposes other than acquiring protective measures. This is not to say Atreya, A., Czajkowski, J., Botzen, W., Bustamante, G., Campbell, K., Collier, B., Ianni, F.,
Kunreuther, H., Michel-Kerjan, E., Montgomery, M., 2017. Adoption of flood
that the Emergency LEAP did not meet its stated goals or objectives. The preparedness actions: a household level study in rural communities in Tabasco,
work of Emergency LEAP cannot be underestimated, however, we Mexico. Journal of Disaster Risk Reduction 24, 428–438.
believe and argue that the program could do better if given a renewed Australian, A.I.D., 2012. Global Facility for Disaster Reduction and Recovery (GFDRR).
Barros, V.R., Field, C.B., Dokken, D.J., Mastrandrea, M.D., Mach, K.J., 2014. Climate
focus that transcends beyond just cash benefits to including better long- Change 2014: Impacts, Adaptation, and Vulnerability. Cambridge University Press,
term interventions like the integration of labour-intensive public works Cambridge.
programs, education, and insurance with the help of relevant stake­ Birkmann, J., Cardona, O.D., Carre~ no, M.L., Barbat, A.H., Pelling, M., Schneiderbauer, S.,
Kienberger, S., Keiler, M., Alexander, D., Zeil, P., Welle, T., 2013. Framing
holders. Community action, instead of household assistance, should be vulnerability, risk, and societal responses: the MOVE framework. Nat. Hazards 67
the bedrock of any program meant to build the resilience of rural (2), 193–211.
communities. Community-driven development (CDD) programs that Brinkhoff, Thomas, 2018. City Population: Ketu South District [Online], Germany, City
Population. Availableat. https://www.citypopulation.de/php/ghana-admin.php?
seek ways of reducing the dismal effects of floods and other natural
adm2id¼0403. (Accessed 10 April 2019).
hazards should be increasingly strengthened for best results. Epa Environmental Protection Agency, 2018. About us: brief history [online], Ghana,
Although the Emergency LEAP does not cater directly for climate- environmental protection agency. Available at: http://www.epa.gov.gh/epa/about.
related risks, it does for disaster victims and some of these disasters Grothmann, T., Reusswig, F., 2006. People at risk of flooding: why some residents take
precautionary action while others do not. Nat. Hazards 38 (1–2), 101–120.
occur as a result of climate change activities. With the existence of Gss Ghana Statistical Service, 2013. 2010 Population and Housing Census: National
Emergency LEAP in Ghana, livelihoods of the vulnerable especially the Analytical Report.
rural populace could also be sustained in the long-term if the govern­ Hallegatte, S., Fay, M., Barbier, E.B., 2018. Poverty and climate change: Introduction, 23.
Environment and Development Economics, pp. 217–233, 3.
ment, implementing ministry, agencies and stakeholders adopt and Herrera, L., Flora, 2017. Climate Change Risk Profile Ghana.
work towards far more better and comprehensive DRR initiatives. Suf­ Kuwornu, J.K.M. (Ed.), 2019. Climate Change and Sub-saharan Africa: the Vulnerability
fice to say that, developing nations can draw a holistic plan aimed at and Adaptation of Food Supply Chain Actors, Wilmington Delaware. Vernon Press.
Mabuku, M.P., Senzanje, A., Mudhara, M., Jewitt, G., Mulwafu, W., 2018. Rural
integrating financial incentives into existing disaster and climate risk households flood preparedness and social determinants in Mwandi district of Zambia
measures through collaborations with respective national agencies and and Eastern Zambezi Region of Namibia. Int. J. Disas. Risk Red. 28, 284–297.
foreign organizations. This collaborative efforts will bring about MESTI Ministry of Environment, Science, Technology, and Innovation, 2013. Ghana
National Climate Change Policy.
renewed focus towards long-term sustainable programs like insurance, NADMO National Disaster and Management Organization, 2016. About us: Ghana
microcredit, public works programs, and comprehensive disaster re­ disaster profile [online], national disaster management organization. Available at:
covery plans. http://www.nadmo.gov.gh/index.php/about-us.
Osberghaus, D., 2017. The effect of flood experience on household mitigation—evidence
We mention two limitations of this study that coud be taken up in
from longitudinal and insurance data’. Global Environ. Change 43, 126–136.
future research. Firstly, we guide the interpretation of our results as Pittman, B., Buta, E., Krishnan-Sarin, S., O’Malley, S.S., Liss, T., Gueorguieva, R., 2018.
correlations as very little attempt is made in causally identifying cash Models for Analyzing Zero-Inflated and Overdispersed Count Data: an Application to
payment impacts. We used different empirical models to confirm the Cigarette and Marijuana Use’, Nicotine & Tobacco Research. official journal of the
Society for Research on Nicotine and Tobacco.
robustness of our estimates but we did not identify the models owing to Poussin, J.K., Botzen, W.W., Aerts, J.C.J.H., 2014. Factors of influence on flood damage
data limitations. Secondly, while the results are interesting and may mitigation behaviour by households’. Environ. Sci. Pol. 40, 69–77.
point to important policy debates in implementing adaptive social pro­ Richert, C., Erdlenbruch, K., Figui� eres, C., 2017. The determinants of households’ flood
mitigation decisions in France - on the possibility of feedback effects from past
tection and disaster risk reduction strategies, the study findings should investments. Ecol. Econ. 131, 342–352.
not be generalized since context matters. However, as most developing Rogers, R.W., 1975. A protection motivation theory of fear appeals and attitude Change
countries are similar from a smallholder household perspective, our 1. J. Psychol. 91 (1), 93–114.
The World Bank, 2013. Social Protection and Climate Resilience. World Bank, Institute of
analysis may provide potential entry and leveraging points for such Development Studies, UKAID, Economic Commission for Africa.
policy initiatives. UNEP, UsNDPS, 2012. Ghana National Climate Change Adaptation Strategy: CC DARE:
Climate Change and Development- Adapting by Reducing Vulnerability.
WACA West African Coastal Areas Management program, 2017. Fighting Coastal Erosion
CRediT authorship contribution statement in Keta Area.

Martin Paul Jr. Tabe-Ojong: Conceptualization, Formal analysis,