Participatory adaptation to climate extremes: an assessment of households’ willingness to contribute labor for flood risk mitigation in Pakistan

Extreme climate events like floods are projected to increase in frequency and intensity in the future (Hirabayashi et al. 2013). Despite several technological advancements, these events are causing significant physical destruction and economic damage to societies (Coumou & Rahmstorf 2012; Kazi 2014). Fluvial flooding is a major type of flood disaster that inflicts large scale damages in developing countries, especially in South Asia, due to the presence of several major river systems (ICIMOD 2007) in the region. In addition, heavy reliance on conventional flood control practices by these nations exacerbates the impacts of riverine floods (Shaw 2006). India, Bangladesh and Pakistan, three neighboring South Asian countries, have suffered heavily due to flooding disasters in last few decades. Despite a long history of flooding, a number of factors render these nations (especially Pakistan) less successful in devising sustainable solutions to manage these extreme events (Kale 2014). An important factor among them is the under-utilization of local knowledge and community resources like labor.

Government bodies, international donor agencies and non-governmental organizations (NGOs) are investing sizeable capital and energy in cushioning the impact of flooding disasters (both as ex-ante and ex-post) but the sustainability of these interventions still remains in question (Pearce 2003; ICIMOD 2007). It is surprising that the possibility of involving the local community (especially in ex-ante measures) in disaster mitigation projects, either physically or financially, is largely overlooked in many developing countries. Involvement of the community in the planning and implementation of disaster mitigation programs adds one more management layer (Nguyen et al. 2011). So, any broad-based participatory approach for disaster management needs to utilize local knowledge and resources (Dekens 2007). Contribution of voluntary labor by flood-prone communities has to be an important aspect of any participatory flood mitigation strategy. As the construction and maintenance of flood defense infrastructure, which are public projects, are funded, partially or solely, by the state, there is a possibility of contribution by users in the form of cash or labor (Chandrasekhar 2005). Thus, knowing about the potential and the extent of in-kind contribution from relatively cash-deprived, flood-prone communities of developing countries would offer valuable insights for disaster management policy and planning.

Generally, flood risk can be reduced by various ex-ante measures, including structural (e.g., building dikes and embankments) and non-structural (e.g., insurance) measures (Grothmann & Reusswig 2006; Kazi 2014). The people's willingness to pay (WTP) for flood insurance to secure themselves against post-flood financial risks has been evaluated in the case of many developed and developing countries (Thieken et al. 2007; Fuks & Chatterjee 2008; Brouwer et al. 2009; Botzen & van den Bergh 2012; Seifert et al. 2013; Abbas et al. 2015) but there is hardly any scientific study that explores people's willingness to contribute (WTC) (in kind) to structural measures against flooding.

Historically, successful intervention programs start and succeed at community level (Laska 1986; Rawlani & Sovacool 2011). Achievement of sustainability of interventions is the core aspect of this approach, as community involvement instills a sense of ownership and attachment to any physical/structural measure aimed at reducing vulnerability to extreme events (Poustie et al. 2014). In addition, utilizing available family labor for flood risk mitigation in developing countries can bring down the costs of structural measures. Most developing countries struggle for the allocation of scarce financial and physical resources among competing ends. This is also true for the allocation of resources for the establishment, repair and maintenance of flood reducing infrastructure (Thunberg & Shabman 1991; Nguyen et al. 2011). The potential contribution of flood-prone communities in building flood protection structures through mobilizing labor is an underexplored research area, though it has been suggested that people who are unable to contribute financially to such measures would be willing to contribute in kind (Brouwer et al. 2009). Nevertheless, Swallow & Woudyalew (1994) and Echessah et al. (1997) report the contribution of labor in the case of controlling the vector of African trypanosomiasis (tsetse fly). In Pakistan, labor contribution for communal structures is not new among communities, as it is already practiced for the brick lining of watercourses (PMU 2005; PMU 2013). So, there is a potential scope for labor contribution in flood prevention structures.

We took insights from the finding of Brouwer et al. (2009) in the case of Bangladesh that the farmers, unable to pay in cash for flood risk mitigation, might be willing to pay in kind, for example, with labor. The existing practice of household labor contribution in irrigation water management in the region including India and Bangladesh (Terpstra 1998; Hussain 2004; Chandrasekhar 2005; PMU 2013), points towards the feasibility of mobilizing labor for flood management. Moreover, the expected increases in frequency and fatalities of flood events necessitate exploring additional options to safeguard communities against these risks given the limited financial capacity of developing countries. With this background, the current study aims at exploring whether the rural communities are willing to contribute their available labor for building new flood protection structures or maintaining the existing ones in order to avoid future damages to their property, crops and livestock. Moreover, the study analyzes the effect of various socio-demographic, economic, physical and behavioral factors underlying the stated labor contribution. This study would also be helpful in designing flood mitigation policy in a developing country context, through providing insights for possible contribution of family labor from communities at risk. Data from the field level survey of flood-affected households by the flooding event in 2010 in Pakistan is utilized to evaluate the scope and extent of WTC labor for flood mitigation and the factors influencing labor contribution decisions.

Before discussing the management of flooding events, we would like to elaborate the geographical context. The Indus River is one of the longest rivers in Asia, with a catchment area of about 1 million km². Sixty five percent of the Indus basin lies in Pakistan, extending to about 75 percent of the country area (Kazi 2014). On its eastern side, the Indus River is joined by major tributaries like the Satluj, Beas, Ravi, Jhelum and Chenab Rivers while the Kabul River is the major western tributary. There are four major causes of flooding in this river basin: (1) heavy monsoon rains in the months of July and August; (2) excessive snowmelt in the Hindu Kush-Himalayan region; (3) natural damming due to landslides or glacial outbursts and (4) westerly waves from the Mediterranean and Arabian seas in winter (ICIMOD 2007; Coumou & Rahmstorf 2012). Heavy snowmelt and torrential monsoons in upstream catchment areas of the Indus Basin (mainly in India) cause flooding in the downstream in Pakistan. The Indus Basin also hosts about 2,500 glacial lakes, out of which 52 lakes could potentially result in Glacial Lake Outburst Floods (Rasul et al. 2011).

Generally, the flood water returns back to the river in the upper reaches of the Indus Basin, but in the lower reaches the flood water does not return to the river resulting in the extended inundation period and associated damages. Embankments or flood protection bunds have been constructed to tame flood waters where over-bank flow is common, although they protect headworks, main barrages and other vital installations. The total length of embankments along major rivers in the country is about 6,719 km (FFC 2009).

Under such geophysical conditions, Pakistan is highly vulnerable to disastrous flooding events. This vulnerability becomes more severe under the changing climatic conditions globally and regionally. In future, the chances of increased frequency, intensity and severity of flooding events are high, as changes in monsoon rainfall patterns in the region could bring intense precipitation events while increased temperatures may also accelerate the melting of Himalayan glaciers (Trenberth 2011).

The mega-flood in 2010 inflicted unprecedented loss on the country. The floods turned more damaging due to lacunas in the disaster risk management plans and policies as well as their poor implementation (Tariq & van de Giesen 2012; Kazi 2014). The disaster management framework of the country largely overlooks ex-ante strategies of managing such natural disasters (Ahmed 2013). Pakistan's National Disaster Management Act (2010) mainly focuses on the mechanisms and activities once the loss has occurred as a result of natural disaster. In contrast, India and Bangladesh have evolved mechanisms to involve the community in the assessment of flood risk, mitigation planning and implementation of flood defense strategy (Shaw 2006; Mirza 2011; Pal et al. 2011; Rawlani & Sovacool 2011). Moreover, Pakistan lags behind India and Bangladesh in terms of protecting its flood-prone population despite having an almost twofold relative vulnerability score compared with its neighboring countries (UNDP 2004). IPCC also highlights poor flood risk governance and weaker institutional coordination in the case of Pakistan compared with improved disaster preparedness in the case of Bangladesh due to proactive community participation and planning process (IPCC 2012).

Ideally, the flood risk management activities follow a cyclic path (Thieken et al. 2007) depending upon the conditions prevailing before and after the flood event. Steps taken at each stage of the cycle would enhance preparedness in the wake of a future event. This is because the steps taken throughout the course of this disaster cycle either help in modifying the causes or minimizing the effects of the disaster on the community, infrastructure and property. Within this cycle, the scope for labor participation looks more promising in building structural flood control measures, leading to increased community preparedness and reduced vulnerability to disaster risk (before the onset of the flooding event) that can potentially reduce the expected damages. Once a flood event has occurred, the community is motivated to protect itself from such an event in the future, whereas governments are obliged to pursue additional measures in safeguarding the community. During this phase, it is possible to involve or at least identify rural labor ready to extend its services.

The integration of readily available rural labor (as the seasonal nature of agricultural labor creates surplus labor in rural areas) into the centrally planned flood mitigation strategy for a specific locality can yield promising outcomes in the long run. Once an effective mitigation strategy has been adapted and implemented in collaboration with the local community using its expertise and resources, people would be motivated to adopt private measures (e.g., flood insurance, building retrofitting) for the protection of their personal property, house and farm enterprise (Thieken et al. 2007).

Flood risk management, being a non-market service with non-guaranteed benefits to a particular individual, needs collective action, one way or the other, thus spreading the associated costs and benefits. Building upon the experience of irrigation water management in a developing country like Pakistan, community participation through labor contribution is envisioned to significantly reduce labor costs in the construction and maintenance of flood-protection structures. For example, in the case of a watercourse lining and cleaning exercise in the country, 100% manual labor was arranged and supervised by community members (Terpstra 1998; PMU 2005). The potential for the replication of a similar exercise for flood risk mitigation seems viable in Pakistan and neighboring countries like India and Bangladesh, as it would also be helpful in exploiting local knowledge and creating a sense of ownership to either newly built or repaired structures. Moreover, it warrants further research in estimating the exact amount of labor required for such flood-protecting structures and assessing the potential for saving labor costs through community labor. This is because all the previously constructed structures are state-built with almost non-existent data on such indicators. Moreover, these structures are too old to accommodate the changed flooding pattern as a result of changing hydrodynamics, socio-geographic conditions and meteorological parameters.

More specifically, the nature of the labor contribution may vary from location to location, but the present study conceptualizes this contribution in the form of manual labor, which may be helpful in constructing or maintaining a flood protection structure with the reduced possibility of depreciation, encroachment or breaching. During the construction phase, local labor can support in earthwork and facilitate technical personnel. The maintenance labor contribution may range from the repair of holes and cuts caused by rain, rodents, cattle, humans and traffic, to turfing, removing of bushes, fine dressing and daily vigilance.

Individuals living in flood prone areas would reduce flood risk at household level by elevating the foundation (plinth) of their houses, installing sand bags and by contributing their labor for building dikes and embankments at community level. Taking insights from previous WTP studies, the amount of labor that a utility maximizing household is willing to contribute for a flood protection scheme, given the perceived probability of future flood event(s), may be influenced by individual characteristics, risk perception, individual preparedness and family resources (Thunberg & Shabman 1991; Zhai et al. 2006; Fuks & Chatterjee 2008; Brouwer et al. 2009; Abbas et al. 2015; Arshad et al. 2015). Households are expected to allocate or contribute a specific amount of labor given the nature and availability of family resources such as the number of adults in the family, household structure (joint or nuclear) and income generation possibilities (farm and off-farm).

WTC labor to community-level flood protection measures may also depend on the costs and labor already invested in terms of individual household (private) measures undertaken (e.g., elevated foundation of the house). Other potential influencing factors could be the distance of the farm from the river and the presence of a flood protection embankment. Studies also show that previous experience of flooding would significantly affect households’ attitudes to precautionary measures against flooding (Thieken et al. 2007; Kreibich et al. 2011). Moreover, the flood experience, along with the amount and nature of damages in previous flood events, may be responsible for influencing the perception of risk associated with a future flood event (Seifert et al. 2013) and hence households’ precautionary options.

We employed a multistage random sampling technique for drawing our sample from the selected districts. A list of all the flood affected villages during the 2010 floods was obtained from each District Government of the selected districts. Five villages from each district were selected at random from this list. A list of flood affected households was obtained from the Numberdar (village head) of each village. There exists no list describing the number of households in a particular village, rather a list of flood affected households assessed for financial compensation, generally based on damage to the house, was prepared by the community leader and officials of the irrigation department and DDMA (District Disaster Management Authority). A sub-list of households who received government assistance of Rs. 25,000 (≈$250) or more was compiled and 10 households were selected at random from this list. This screening was carried out to ensure proper representation of flood victims with reasonable flood damage. Only the household head was contacted to seek information, but in his absence at the time of the survey, the most senior family member was requested to respond to the questionnaire. All the selected households agreed to participate in the survey except those not present at home at the time of the personal interview (about 10 households). Accordingly, these households were replaced by the same procedure as that of the initial selection. The responding family head was informed to represent his family and convey the amount of family labor ready to be used for a flood protection scheme.

In total, 250 respondents from the selected villages were interviewed using a structured and pre-tested questionnaire. The questionnaire for this study sought information about socio-demographic background, flooding profile and flood impacts during the 2010 floods, and elicitation as to the WTC family labor in the construction or repair/maintenance of a flood protection structure. This elicitation was based on a hypothetical labor contribution for the construction of a flood-mitigating embankment using the Contingent Valuation Methodology (CVM). This elicitation procedure is used by many studies dealing with WTP responses in the case of future proposed intervention programs (Thunberg & Shabman 1991; Echessah et al. 1997; Zhai et al. 2006; Fuks & Chatterjee 2008; Zhai & Suzuki 2008; Akter et al. 2009; Botzen & van den Bergh 2012). Although people of the study area are aware of the labor contribution for the watercourse lining and cleaning program, there exists no such mechanism for the contribution of labor for the repair or construction of flood protection structures. Using CVM techniques, this study tried to unravel the potential labor contribution among the flood affected households of the study area. Nevertheless, a potential limitation of the study may be the sample size, which is justified by the financial and time constraints faced by the research team.

During the survey, respondents were briefed about the whole mechanism of labor participation and the resulting benefits from this exercise. This briefing also included the information about the nature and timing (e.g., once or in multiple episodes) of labor contribution and possible benefits to the community (in terms of reduced risk of flooding) and support to the field staff. Many households were aware of the benefits of contributing labor towards common structures, especially in water management on agricultural lands. One such example is the brick lining of watercourses started in 2004, where full labor (or its cost) is being contributed by the community (PMU 2005). Historically, the maintenance of older, unlined watercourses was also done bi-monthly through community participation (Terpstra 1998).

The actual data collection was carried out during October–December 2012. We employed the face-to-face mode for the survey due to its suitability to the local conditions and target population, keeping in view the poor educational background and physical infrastructure such as poor internet and postal facilities. Households were informed explicitly about the purpose of this exercise and no financial or physical incentives were offered or promised.

In the survey, households were asked to state their preferences in terms of labor that they are ready to contribute for a flood mitigating structure to reduce future flood related losses. This type of elicitation is sufficiently embedded in CVM being a survey-based procedure for valuing benefits from changes in non-market commodities (Thunberg & Shabman 1991). Though there are a handful of CVM studies that determine WTP for flood protection measures in developing countries, the use of this technique is less common for payments in kind (e.g., labor, materials).

The aim of the WTC exercise was to elicit a range of man-days a household would be willing to contribute annually in a flood alleviation scheme to reduce the future flood risks. This study used 11 hypothetical bids for labor contribution (Table 1). These bids were developed carefully, keeping in view the past experience of labor participation in other water management strategies (Terpstra 1998; PMU 2005). In addition, pre-testing from 10 households and expert opinion significantly guided the development of these bids (the pre-tests are not part of the analysis). The respondents were briefed about the relevance and practicability of the concept of participatory flood management through labor contribution. This could greatly reduce the hypothetical bias and hence reduce, to a great extent, the perceived skepticism about whether the contingent scenario would take its effect in reality. We used the double-bounded dichotomous choice CVM (DB-DCCVM) model for analyzing WTC. At the first step, the initial bid (IB) (in man days) was randomly assigned to the respondents to avoid starting point bias, and respondents were asked to accept or reject the bid. An alternative follow-up higher bid in cases of positive response, or a lower bid in cases of negative response to IB, was then offered. These follow-up bids were double or half the IB depending upon a positive or negative response, respectively, to the IB (Kitamura et al. 2009). Respondents were free to accept or reject these follow-up bids.

For the determination of the confidence interval of the WTC, we resorted to the Krinsky and Robb procedure (Park et al. 1991). This step is necessary because parameter estimates γ and δ are also random variables. We predicted the confidence interval for WTC using the information contained in the estimated DB Logit model. To calculate the confidence interval, a new parameter vector is generated by multiple random draws from assumed multivariate normal distribution having mean and variance-covariance vector ⁠. WTC is then calculated using each drawing of ⁠. Using the complete set of replications, an empirical distribution of WTC is obtained and then ranked. Dropping γ/2 values from each tail of the ranked distribution, a (1 – γ) confidence interval, called the Krinsky-Robb confidence interval, is obtained.

Methodologically, DB-DCCVM is advantageous over a single-bounded dichotomous choice CVM in that it can yield a range of responses of people's willingness even with a small sample size. Moreover, by the use of follow-up bidding procedure, it is helpful in addressing the issue of misspecification of IB, anchoring effect and passively ‘yeah’ saying.

Socioeconomic characteristics of households could influence the perception and preparedness for a disaster risk (Kellens et al. 2011). Socio-economic characteristics of the surveyed households are presented in Table 2. The average age of the respondents, who in most cases are represented by the family head, is about 53 years. This shows an aging farming class of rural families in the study area. The average number of schooling years of the household head is around 5 years, while the average family size (only adults) is about 6 members. The contribution of labor for flood protection measures is also hypothesized to depend on the nature of the family system in the study area. It is interesting to note that around two-thirds of the sampled households belong to the joint family system.

About 25% of the sampled households have cemented-brick houses and 25% have mud houses, while the rest of them have houses made of kiln-processed bricks fixed with mud. This aspect is relevant, as the people having firm building structures might be less willing to contribute labor. The opposite may also be true, as such households would be more interested to save their heavy investment in the building of a house with costly materials. Another interesting aspect is the land ownership pattern. The mean owned land and rented-in land area are about 2.48 ha and 1.13 ha, respectively. Average farm distance from the river is 2.78 km, while the average plinth of the house is about 1 meter. The average monthly total income of the surveyed households is Rs. 25,971. About 55% of total income comes from farming activities, whereas 45% percent is contributed by off-farm activities such as Government or private service, fishing, a business or shop, tailoring and day labor. The average amount of relief and reconstruction support received by the sample households is Rs. 62,864. A major portion of this compensation is contributed by the provincial government with some contribution from NGOs and federal government.

Respondents were offered random bids of labor contribution. Their response to these bids varied with respect to various socio-demographic characteristics. About half of the respondents (44%) show their WTC labor for a flood mitigation project, though about 31% of respondents accepted the subsequent lower bids (Table 3). This shows the potential for participation of labor in the study area to contribute to the establishment and maintenance of flood protection structures. The percentage of the households declining both proposed bids for labor contribution also shows that they were aware of the resource constraints and were not passively answering the hypothetical question.

The percentage distribution of respondents according to the number of flood events experienced and their response to the contribution of labor is presented in Table 3. The number of respondents who made at least one ‘yes’ response declined rapidly with exposure to a third or fourth flood event. This attitude is best explained by their enhanced preparedness, based on experience of previous disasters, due to increased awareness about future risk and associated losses forcing them to take additional measures against flooding (Joerin et al. 2012). The fatalistic attitude on the part of more experienced households is somewhat non-existent as there exists a positive association between number of floods experienced and the degree of belief in the efficacy of proposed community level flood management measure as shown by the correlation measure of Spearman's rho (ρ = 0.205).

Apart from flood experience, family structure (joint or nuclear), the number of adult family members and the perception about the effectiveness of the proposed project may have some influence on labor contribution decisions as well (Kellens et al. 2011). Labor contribution decisions and hence adaptation actions may vary if the perception of households about the future proposed scheme is positive or negative (Lindell & Hwang 2008; Mishra et al. 2010). Table 4 shows that out of the total surveyed joint families, about half of them show their WTC labor, whereas more than half of the nuclear families declined any contribution choice. The joint family system prevails in most of the rural areas of Pakistan, where more than one married couple along with their siblings reside within one boundary wall. Moreover, joint families have surplus labor or they can rationalize labor allocation to flood alleviation projects. Results show that a fraction of nuclear families opted for the follow-up lower bid (23.4% of the total households opting for follow-up lower bids among nuclear families). These findings are further confirmed when we look at the rural families’ categories based on the number of adults in the households. About 75% of the small families do not show any WTC labor for flood protection schemes, while the medium- and large-sized families show relatively higher WTC. There is no significant difference in the response percentage of households belonging to medium and large families.

Results in Table 4 also suggest that positive perception about the proposed measure affects the labor contribution decisions by the households. This is in line with the findings of Lindell & Hwang (2008) and Seifert et al. (2013), who posit that the perception about the effectiveness of any future mitigation project plays a vital role in the shaping of adaptation and mitigation decisions. More than half of the sampled households who believed in the effectiveness of the proposed scheme were willing to contribute labor. The percentage of households who did not believe in the effectiveness of such structures involving community laboring was about 23% of the total sample, and only 12.1% of those who opted for the lower bid option (Table 4) were willing to contribute labor, reflecting their reluctance to contribute labor.

It is hypothesized that socio-demographic, physical, economic and behavioral attributes influence the WTC decisions of households. The following section discusses the estimated regression coefficients of the double bounded Logit model. Most of these coefficients exhibit expected a priori relations to WTC (Table 5).

Social and demographic characteristics influence the decision making process of households to a considerable extent. The age of the household head has a significant and negative influence on the probability of accepting a labor contribution choice for flood protection work. This finding implies that the probability of saying ‘yes’ to a labor contribution choice decreases with an increase in the age. This may be due to the fact that older people may have lower trust in collective action, due to past negative experiences of free-riding behavior of some community members. It may also be due to their enhanced knowledge about the possibilities of flood management at household level. The coefficient for education of the household head has a negative and significant impact on WTC decisions. A well-educated rural family head is generally aware of the significance of education, and hence there is increased probability of more educated family members (Khan & Ali 2003). Therefore, rural families with more education might have higher opportunity costs and less free labor. These findings also suggest that naive, less educated respondents are more willing to contribute family labor for the construction of flood protection structures. On the other hand, one can expect a free-riding intent on the part of more educated households exhibited by their bidding behavior. This is supported by the findings of Yan et al. (2014) who show a significant positive impact of education on the free riding index in the case of youth and middle age groups.

The coefficient on labor contribution bid offered to the respondents has an expected negative sign and is statistically significant. This implies that the probability of saying ‘yes’ declines with the increasing level of labor contribution demand (bid). This finding is consistent with the findings of Fuks & Chatterjee (2008) and Akter et al. (2009) eliciting household responses for WTP for a flood control project. Although the coefficient for joint family structure has a positive sign, it is not statistically significant. On the other hand, the coefficient for the number of adult family members has a statistically significant value, which implies that as the number of adult members in a family increases, the probability of contributing labor for flood mitigating schemes would also increase. This is justifiable, as more adult members in a family are a source of extra and useful labor that can effectively contribute to such schemes.

In the case of physical attributes of the surveyed households, only the distance from the river has a statistically significant influence on WTC decisions. The coefficient of distance from the river has a negative sign, as expected, and implies that the greater the distance from the river, the lesser will be the probability that a household would opt to contribute labor for a flood protection scheme. An indirect but closely related finding by Lindell & Hwang (2008) supports our result. They show that perceived personal risk is significantly correlated with proximity to a water body. This perceived personal risk may motivate individuals to take disaster protection decisions. Zhai et al. (2006) also show a negative and significant impact of distance from the river on WTP for flood risk reduction. The negative relation of WTC to distance of the farm from the river is expected as it reduces the severity of flood damage and instills a sense of security (Brouwer et al. 2009). The variables for house type, owned land area and rented land area have statistically non-significant values, implying that they do not affect WTC decisions significantly. The presence of the embankment has a positive and non-significant coefficient.

Farmers were asked about their farm and off-farm income to evaluate their role in the WTC decisions. The coefficient for farm income is negative and significant. It implies that an increase in the farm income would reduce the probability of labor contribution in a flood alleviation scheme. This is in accordance with our expectations, as the farmers with increased farm income would find it less attractive as the perceived opportunity cost of labor is increased. Although increased farm income is an indication of a larger farm, it still implies more engagement of family labor at the farm, and the household head may overvalue and find it unattractive to spare labor for collective purposes. In addition, farmers with higher income may feel they are destined with enough resources to deal with potential damages from flooding.

The coefficient of livestock damage in a previous flood event significantly and positively affects the probability of saying yes to a specific labor contribution choice. On the other hand, other flood damage indicators do not show a significant impact on WTC choices. This is because the loss to crops and house cannot be avoided during a flood event as they are immovable. Moreover, farmers in the study area can pursue additional measures to safeguard livestock, such as moving them to higher places or transporting them to safer locations. Nevertheless, an increased amount of loss to livestock in a sudden flood disaster could potentially be avoided by erecting some sort of flood protection embankment, either by the community itself or with support from the government. The precautionary attitude driven by the extent of damage in the previous disastrous events (Seifert et al. 2013) could be the explanation for the observed relationship between WTC labor and livestock damages.

The amount of compensation received in a previous flood event has a significant and positive effect on labor contribution decisions. Against our expectation, this finding seems plausible. It implies, on one hand, that an increased amount of compensation would persuade farmers to work more, i.e., to offer more labor for flood protection; but it also implies, on the other hand, increased risk perception due to heavy flood damage in a previous flood event (a higher compensation indicates higher flood damage). The argument by Kreibich et al. (2011) validates this finding to the extent that complete governmental relief disincentivizes private precautionary actions, while setting limits on compensation would motivate people to take preventative actions. Specific to the study area, compensation to flood victims is generally based on the assessment of visible losses to their house and livestock, with little consideration to crop losses. This process, too, is laden with many bureaucratic hurdles, underhand practices and untimely payments. Therefore, it is a testimony to the fact that sample responses are not distorted or biased due to expected compensation, otherwise they could refuse to offer any sort of contribution as they expect the damages will be covered by the compensation.

The coefficient on the perception about the effectiveness of the proposed scheme is statistically significant and positive, as expected. It implies that the probability of a ‘yes’ to labor contribution choices significantly increases with the increase in a household's confidence in the usefulness of the proposed measure. This finding is in line with the findings of Grothmann & Reusswig (2006), who argue that people take those precautionary measures which they believe to be effective in protecting them against flood risk. Moreover, as the belief or confidence in an intervention is affected by the previous outcome of similar intervention, the possibility of taking similar decisions increases (Eiser et al. 2012). The familiarity and confidence of the sample farmers in the practice of labor contribution, and allied benefits in irrigation water management, seem to play a decisive role in agreeing to the proposed intervention for flood risk mitigation.

The estimated mean labor contribution elicited through WTC exercise is 11.07 man-days/year per household (Table 6). This finding affirms the proposition of the possible contribution of labor put forth by Brouwer et al. (2009). The elicited amount of labor in this case study represents a significant contribution in lowering labor cost for the construction and maintenance of a flood protection structure. The yearly saving would amount to Rs. 4,084 per year per household (considering the average daily wage rate of rural workers to be equal to Rs. 369 (GoP 2013)). In the recent past, the labor contribution by the local community has also proved to be significant in the case of brick-lining of watercourses. Under this program, the entire labor cost (mostly in kind) was contributed by the farming community, amounting to 36% of the total cost. This also included the labor required for the demolition of the old unlined and katcha (earthen) watercourse and payment for masons (PMU 2005, p.66). Contribution of labor in the case of cleaning and lining the katcha watercourses has been in place for decades through communal action, with technical and material (partial) support from the government (Terpstra 1998, p.77).

The contribution of labor for a structural flood control structure in a developing country seems plausible due to two main facts. Firstly, the disaster prone people prefer to eliminate the risk using structural measures (Botzen et al. 2013) rather than to minimize it through other soft measures. Secondly, given surplus labor in the agriculture sector of a developing country and the absence of any flood insurance mechanism or lower capacity to pay premiums, farmers would be inclined towards such means of resourcing structural protection measures. In addition, if present, the insurance companies may find it unattractive to offer their services to those farmers with poor or little structural defense measures against flooding, or the compensation would fall short of the losses that they could incur from a flooding event.

The study revealed the willingness of rural households to protect themselves from flooding events by contributing labor for structural flood protection measures such as embankment and dikes. However, the decision to contribute labor is largely influenced by the socioeconomic, physical and behavioral characteristics of households. These include the level of labor contribution requested, the age and education of the household head, belief in the proposed scheme in alleviating flood risk, distance from the river, farm income and damage to livestock in previous flood events, and governmental relief received in a previous flood event. The WTC labor may increase with the number of adult family members, loss of livestock and the amount of compensation received in previous flood events as well as enhanced confidence in the proposed scheme. But it may decrease if the distance of the farm's location from the river is larger; the farm income is higher; the labor contribution requested is bigger; and the age and education of the family head are higher. The estimated mean contribution of labor is expected to reduce the building and maintenance costs of the flood protection structures, supported by the evidence from past efforts in participatory irrigation-infrastructure development and maintenance (PMU 2013). The expected reduction in the cost due to rural labor's contribution amounts to Rs. 4,084 per household/year (≈39 USD). This contribution is possible and achievable given the seasonal nature of agricultural labor, which generates surplus labor. Therefore, integration of community participation through labor contribution within flood mitigation plans in developing country like Pakistan would ensure sustainability of the measures with reduced costs. Moreover, linking governmental relief efforts with private initiatives and documented labor contribution in common structures like embankments or dikes can be helpful in minimizing private property damages, and can reduce claims for compensation through public funds.

The current study had financial and time constraints, and hence there is scope to further investigate the exact amount of labor required for constructing flood protection embankments and assessing the share of contribution from the community. In addition, the effect of age and education of the household head on labor contribution decisions can be explored further. It will also be interesting to see how the free-riding behavior of community members in earlier collective-action exercises affects the WTC.

The authors would like to thank the financial assistance provided by Higher Education Commission of Pakistan (HEC) through German Academic Exchange Service (DAAD) for the first author. The financial support for the survey provided by Stiftung Fiat Panis is also highly encouraged.