What Influences the Demand for a Potential Flood Insurance Product in an Area with Low Previous Exposure to Insurance? – A Case Study in the West African Lower Mono River Basin (LMRB)

Wagne , Simon; Thiam, Sophie; Dossoumou, Nadège I. P.; Daou, Da id A icle — Published Ve sion Wha In luences he Demand o a Po en ial Flood Insu ance P oduc in an A ea wi h Low P e ious Exposu e o Insu ance? – A Case S udy in he Wes A ican Lowe Mono Ri e Basin (LMRB) Economics o Disas e s and Clima e Change P o ided in Coope a ion wi h: Sp inge Na u e Sugges ed Ci a ion: Wagne , Simon; Thiam, Sophie; Dossoumou, Nadège I. P.; Daou, Da id (2023) : Wha In luences he Demand o a Po en ial Flood Insu ance P oduc in an A ea wi h Low P e ious Exposu e o Insu ance? – A Case S udy in he Wes A ican Lowe Mono Ri e Basin (LMRB), Economics o Disas e s and Clima e Change, ISSN 2511-1299, Sp inge In e na ional Publishing, Cham, Vol. 8, Iss. 1, pp. 1-32, h ps://doi.o g/10.1007/s41885-023-00138-w This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/318348 S anda d-Nu zungsbedingungen: Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen Zwecken und zum P i a geb auch gespeiche und kopie we den. Sie dü en die Dokumen e nich ü ö en liche ode komme zielle Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich machen, e eiben ode ande wei ig nu zen. So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen (insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en, gel en abweichend on diesen Nu zungsbedingungen die in de do genann en Lizenz gewäh en Nu zungs ech e. Te ms o use: Documen s in EconS o may be sa ed and copied o you pe sonal and schola ly pu poses. You a e no o copy documen s o public o comme cial pu poses, o exhibi he documen s publicly, o make hem publicly a ailable on he in e ne , o o dis ibu e o o he wise use he documen s in public. I he documen s ha e been made a ailable unde an Open Con en Licence (especially C ea i e Commons Licences), you may exe cise u he usage igh s as speci ied in he indica ed licence. h p://c ea i ecommons.o g/licenses/by/4.0/ Vol.:(0123456789) Economics o Disas e s and Clima e Change (2024) 8:1–32 h ps://doi.o g/10.1007/s41885-023-00138-w 1 3 RESEARCH Wha In luences heDemand o aPo en ial Flood Insu ance P oduc inanA ea wi hLow P e ious Exposu e oInsu ance? – ACase S udy in heWes A ican Lowe Mono Ri e Basin (LMRB) SimonWagne 1,2· SophieThiam3· NadègeI.P.Dossoumou4· Da idDaou2 Recei ed: 12 Augus 2023 / Accep ed: 19 No embe 2023 / Published online: 18 Decembe 2023 © The Au ho (s) 2023 Abs ac Floods po ay a se e e p oblem in he i e ine a eas o Wes A ica while mo e equen and in ense hea y p ecipi a ion e en s a e p ojec ed unde clima ic change scena ios. Al eady, loods cause mani old impac s, lea ing he popula ion o cope wi h he inan- cial impac s o loods h ough hei own means. As o mal isk ans e mechanisms (e.g., insu ance) a e no ye widely a ailable o he popula ion, e o s o inc ease hei acces- sibili y a e being in ensi ied. Howe e , s udies assessing lood insu ance demand cu en ly mos ly ocus on egions wi h mo e es ablished ma ke s. Also, hey a e majo ly applying con en ional s a is ical modeling app oaches ha conside only a small numbe o pa am- e e s. Con a ily, his s udy aims o p o ide an app oach o assessing lood insu ance in a con ex o low p e ious exposu e o such p oduc s, o allow o a be e conside a- ion o he esea ch con ex . The e o e, a pa ame e selec ion amewo k is p o ided and machine lea ning and deep lea ning models a e applied o selec ed pa ame e s om an exis ing household su ey da a se . In addi ion, he deep lea ning sequen ial neu al ne - wo ks ou pe o med all machine lea ning models achie ing an accu acy be ween 93.5— 100% depending on he loss unc ion and op imize used. The isk o be co e ed, insu ance pe cep ion, no access o any sou ce, access o suppo om communi y solida i y unds, access o go e nmen al suppo , o d awing upon own esou ces o inancial coping, inan- cial eco e y ime, lack o means and p io i izing mo e essen ial needs eme ged as impo - an model pa ame e s in esea ching insu ance demand. Fu u e oll-ou campaigns could conside he pa ame e s poin ed ou by his s udy. Keywo ds Floods· Machine lea ning· Deep lea ning· Willingness o insu e· Togo· Benin In oduc ion O e he pas decades, he e ha e been obse a ions o an inc easing end o hyd ological ex emes (i.e. maximum peak discha ge) in Wes A ica, leading o an inc ease o disas- ous lood e en s in a eas loca ed in p oximi y o la ge i e s (Ranasinghe e al. 2021). Ex ended au ho in o ma ion a ailable on he las page o he a icle 2 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Mo eo e , while o e all p ecipi a ion is p ojec ed o dec ease in Wes A ica, hea y p e- cipi a ion e en s a e expec ed o occu mo e equen ly and in ensi ely acco ding o sce- na ios conside ing medium o high emission le els, which leads o accumula ed hyd o- clima ic s ess h ough d ough and lood e en s in he egion (T isos e al. 2022; Gio gi e al. 2019). Al eady, loods cause a wide a ie y o impac s in Wes A ica, such as dam- aged buildings, dis up ion o li elihoods, damaged goods, a ali ies, displacemen , sick- ness and sp eading o diseases, damaged in as uc u e and c op damage (Wagne e al. 2021; A iyie e al. 2018; B isibe and Pepple 2018; Addo and Danso 2017; Ahadzie e al. 2016; Ene e e al. 2016; Adewole e al. 2015; Adelekan and F egene 2015; Codjoe e al. 2014). Wi h ega ds o he inancial implica ions o lood impac s in he Lowe Mono Ri e Basin (LMRB) in pa icula , i was ound ha loods egula ly a ec households inancially h ough ag icul u al (los in es men s h ough loss and des uc ion o c ops and plan a- ions, loss o li es ock), ma e ial ( epai and eplacemen cos o damage o des uc ion o esiden ial houses and pe sonal ma e ial belongings), heal h (sickness and subsequen pay- men o medical ca e), and comme cial/ ade impac s (los income om damaged s o ed p oduc s o sale, lack o ma ke access, and a ec ed ma ke places) (Wagne e al. 2022). While mu ual suppo among a ec ed households, especially in he phases o esponse and econs uc ion (especially hos ing lood ic ims and helping neighbo s o ebuild) (Lamond e al. 2019; Amoako e al. 2019; Ahadzie e al. 2016; Codjoe and Issah 2016; Adelekan and Asiyanbi 2016), seems o be e y p e alen in he Wes A ican egion, he e appea s o be a lack o isk ans e ins umen s ha a e designed o add ess he inancial consequences o loods (Wagne e al. 2021). Thus, people in he egion equen ly eso o in o mal mechanisms ha a e no o iginally designa ed o alle ia ing he di e se inancial implica- ions o lood impac s, which se s households back in hei inancial achie emen s (Wagne e al. 2022; Boubaca e al. 2017; Addo and Danso 2017). Mo eo e , he equency and se e i y o lood impac le els in he LMRB equi e mo e conce ed isk educ ion ac i i ies be o e es ablishing isk ans e mechanisms, such as insu ance, ha enable sp eading he isk o inancial losses ac oss a la ge pool o bene i- cia ies (Wagne e al. 2022). Also, whe he insu ance is an app op ia e isk managemen ool in de eloping economies o no emains a con es ed issue (Pill 2022; Mechle and Deubelli 2021; Dehm 2020; Linne oo h-Baye e al. 2019; Schä e e al. 2019; Gewi zman e al. 2018). While he e a e inc eased e o s o aise insu ance pene a ion and insu ance co e age agains clima e- ela ed ex eme e en s in de eloping economies (InsuResilience Global Pa ne ship 2021), insu ance p o ec ion agains lood impac s emains di icul o be es ablished, e en globally (Lége 2022; Flood Resilience Ini ia i e 2020; Lloyd’s 2018). In addi ion, much o he esea ch on he up ake o o willingness o pay o lood insu ance ocusses on he Asian, No h Ame ican and Eu opean egion, in which he es ablishmen o lood insu ance in he ma ke and amilia i y wi h such p oduc s a e e y di e en om he Wes A ican egion. Aside om a ew s udies (Be g e al. 2022; Oduniyi e al. 2020; Na ud and Vondolia 2020; Adzawla e al. 2019), his opic has no been widely esea ched in he A ican con ex . Also, insu ance pene a ion on he A ican con inen in gene al is only hal o he global a e age while also he a e age p emiums pe pe son a e ele en imes lowe (Bagus e al. 2020). Thus, o be e in o m u u e oll-ou campaigns o lood insu ance p oduc s i is impo an o esea ch he pa ame e s ha a e associa ed wi h insu - ance ake-up in se ings whe e a la ge numbe o people a isk ha e no ye been insu ance cus ome s, such as he LMRB. Mos s udies esea ching he willingness o insu e (WTI) agains loods/willingness o pay (WTP) ely on pa ame e selec ion di ec ly based on li e a u e and subsequen ly apply eg ession me hods (Ne usil e al. 2021; Robinson and Bo zen 2019; Reynaud e al. 2018; 3 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Fahad and Jing 2018; Tu ne e al. 2014; Bo zen e al. 2013, Bo zen and an den Be gh 2012), ha usually only conside a low numbe o pa ame e s. Con a ily, i p esen s a challenge o de i e such pa ame e s om a conside able body o s udies o he Wes A i- can egion, due o he limi ed numbe o a ailable publica ions om his a ea. Thus, es ab- lished amewo ks o easons o pa ame e inclusion om o he con ex s migh no be he bes i ing o his esea ch con ex . To add ess his gap, his s udy in es iga es he ollow- ing cen al esea ch ques ion: Which pa ame e s in luence he decision-making p ocess o households o ake up a po en ial insu ance p oduc agains lood damages in a se ing wi h low p e ious exposu e o such p oduc s, such as he LMRB? Cons ained by he limi ed li e a u e base o he Wes A ican egion, his s udy ini- ially e iews li e a u e on WTI agains loods/WTP o lood insu ance on a global scale. Based on his body o li e a u e, a amewo k is de eloped ha summa izes six hema ic a eas o pa ame e s (subjec i e pe cep ion o lood isk, objec i e lood isk, in e ac ions wi h insu ance ins i u ions, In e ac ion wi h o he ins i u ions & social en i onmen , a ib- u es o HH/indi iduals, asse s o be po en ially insu ed) o guide which ac o s a e in luen- ial on he demand o insu ance in he esea ch se ing. To s uc u e he pa ame e selec- ion, ea u e columns o he en i e da a se we e ini ially assessed o he en i e da a se . Then, he emaining pa ame e s we e ca ego ized in o he six hema ic a eas o he ame- wo k. Mo eo e , he g ouped pa ame e s we e assessed h ough pai plo s and a hea map co ela ion ma ix. As a inal s ep o e i ica ion, c oss abs we e used o assessing he co - ela ion be ween he pa ame e s and he ou pu alue. This da a-d i en pa ame e selec ion app oach is deemed sui able o his s udy due o esea ching a con ex in which people a isk ha e no been widely exposed o insu ance p oduc s. Subsequen ly, on he basis o he selec ed pa ame e s, machine lea ning and deep lea ning models a e ained ha se e in explaining he obse ed demand o a po en ial lood insu ance p oduc in he esea ch a ea. Backg ound Insu ance andRisk T ans e o Floods inTogo andBenin Cu en ly, insu ance p oduc s agains he impac s o loods a e no widely o e ed on a household le el in Togo and Benin. The insu ance indus y is mos ly cen e ed a ound mo o cycle/ca insu ance and less on na u al haza ds (Me on 2019). In addi ion, he e a e e o s in Benin o es ablish heal h insu ance in pilo communi ies ee o cha ge o i s bene icia ies in he i s h ee yea s (Go e nmen o he Republic o Benin 2021). Wi h ega ds o loods, calls o a easibili y assessmen o a lood insu ance sys em h ough a na ional insu ance und a e e en da ing back o a leas 2011, as s a ed in a pos -disas e needs assessmen o he 2010 loods (Go e nmen o he Republic o Benin 2011). Also, he Togolese go e nmen exp essed a s ong in e es in easibili y s udies o an ag icul- u al insu ance sys em wi hin i s Na ional Adap a ion Plan (Go e nmen o he Republic o Togo 2017). In addi ion, in 2018 Togo was chosen by he pan-A ican isk pool mecha- nism A ican Risk Capaci y (ARC) o se e as a pilo coun y o he implemen a ion o a lood insu ance scheme (Akoda 2018). Howe e , no in o ma ion on i s cu en s a us could be ound, and he mos ecen a ailable epo o he Togolese Republic only con ains in o ma ion o he e en o d ough (A ican Risk Capaci y 2021b), simila ly o Benin (A ican Risk Capaci y 2021a). Mo eo e , he Beninese go e nmen also s a ed a p ac ical 4 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 absence o an insu ance sys em o clima e- ela ed impac s, such as loods, d ough s, wind s o ms, o hea wa es, despi e hei po en ially high impac on he coun y’s g oss domes ic p oduc (Go e nmen o he Republic o Benin 2020). Rega ding he LMRB in pa icula , a ecen s udy poin s ou a s ong need o isk- educing lood adap a ion measu es and ha a con en ional, ma ke -based lood insu ance app oach could be imp ac ical due o he high se e i y and equency le els o epo ed lood impac s om a household pe spec i e (Wagne e al. 2022). As a consequence, his s udy aims o show ele an insigh s in o he po en ial lood insu ance ma ke , o he case ha isk- educing lood adap a ion measu es a e success ully implemen ed in he LMRB. Mo eo e , he esea ch p o ides insigh o insu e s o see i hey could help o opening a ma ke o hemsel es by con ibu ing o in es ing in o lood adap a ion measu es in he a ea. Finally, his esea ch could bene i he p e iously men ioned endea o s o es ablishing lood insu ance ha a e al eady aking place and suppo hei po en ial ollou campaigns. S udies Resea ching heDemand o Flood Insu ance Va ious s udies on he demand o insu ance and hei in luen ial ac o s ha e been pub- lished in he pas yea s unde he ields o willingness o pay (WTP) o willingness o insu e (WTI). Whe eas he o me s ide is mainly ocusing on calcula ing a p emium ha po en ial insu ance clien s a e willing o pay, he la e usually esea ches he gene al in e es le el among a ge ed g oups. The la e aspec also po ays he main ocus o his s udy. Howe e , only a small numbe has esea ched he in luen ial ac o s on demand o lood insu ance in he A ican con ex (Be g e al. 2022; Oduniyi e al. 2020; Na ud and Vondolia 2020; Adzawla e al. 2019). The majo sha e o s udies om ha s ide o esea ch ocused on he Asian (Hossain e al. 2022, Senapa i 2020a, b, Liu e al. 2019, Dewi e al. 2018, Reynaud e al. 2018, Sidi e al. 2018, Fahad and Jing 2018, A shad e al. 2016, Ren and Wang 2016, Abbas e al. 2015, Aliagha e al. 2015, Aliagha e al. 2014, Tu ne e  al. 2014, Hung 2009), No h Ame ican (Da ling on and Yiannakoulias 2022; Huang and Lubell 2022; Ne usil e al. 2021; This le hwai e e al. 2020; A eya e al. 2015; Oulahen 2015; Kousky 2011; B owne and Hoy 2000) o Eu opean con ex s (Osbe ghaus and Rei 2021; Robinson and Bo zen 2020, 2019; Bo zen e al. 2013; Sei e e al. 2013, Bo zen and an den Be gh 2012) – a eas in which lood insu ance sys ems and insu ance in gene al a e mo e widely es ablished. In s udies om his s ide o esea ch, he in luen- ial ac o s men ioned ha e o en been g ouped in o di e en ca ego ies o p o ide be e o ien a ion o esea che s in he selec ion o ele an pa ame e s (summa ized in Table1). Fo example, Sei e e al. (2013) s a e he in luence o pe cep ions o lood isks (subjec- i e iews), expe iences wi h lood impac s (objec i e iews) as well as ac o s ela ing o in e ac ions wi h disas e assis ance om ins i u ions (humani a ian/public compensa ion). Simila ly, Ne usil e al. (2021) also poin ou he impo ance o ac o s exp essing subjec- i e and objec i e iews on lood isk, while adding he cha ac e is ics o esiden ial houses (asse s) and demog aphic cha ac e is ics o he esponden s (a ibu es o HH/indi idual). Aliagha e al. (2014) as well aise he in luence o objec i e and subjec i e iews on lood isk and socio-economic/demog aphic ac o s. To achie e i s objec i e, his s udy compiles u he in luen ial ac o s om u he WTP/WTI s udies om a global scope/ a ious geo- g aphical con ex s and g ouped hem as well in o dis inc ca ego ies while d awing upon and complemen ing he sugges ed ca ego ies om he p e iously men ioned s udies. In ha way, a amewo k o suppo he selec ion o in luen ial ac o s was c ea ed o his s udy (Fig.1). 5 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Table 1 Summa y o pa ame e s men ioned in WTP/WTI s udies Ca ego y Thema ic a ea Pa ame e Re e ences Compa able pa ame e in su ey da a se Flood isk “Subjec i e” pe cep ion o lood isk Flood isk pe cep ion (Hossain e al. 2022, Reynaud e al. 2018, Oulahen 2015, Sei e e al. 2013, Bo zen and an den Be gh 2012, Hung 2009) Yes Recen ly) expe ienced lood e en s (Osbe ghaus and Rei 2021; Senapa i 2020a; Liu e al. 2019; Adzawla e al. 2019; Fahad and Jing 2018; Ren and Wang 2016; A eya e al. 2015; Aliagha e al. 2014; Tu ne e al. 2014; Hung 2009; B owne and Hoy 2000) Yes Pe cep ion on clima e change (Adzawla e al. 2019; Oulahen 2015, Bo zen and an den Be gh 2012)Yes Awa eness (Senapa i 2020b)Yes An icipa ed wo y and eg e abou uninsu ed losses (Robinson and Bo zen 2020, 2019)Yes The obse a ion o o he ’s losses (Tu ne e al. 2014)Yes “Objec i e” Flood Risk (Ex e nally de ined) le el o lood isk (Huang and Lubell 2022; Ne usil e al. 2021; Kousky 2011)Yes P oximi y o i e s (Sidi e al. 2018, Bo zen and an den Be gh 2012, Kousky 2011) Indi ec ly con ained in o he pa ame e o lood isk Li ing in a low lying a ea (Bo zen and an den Be gh 2012) Indi ec ly con ained in o he pa ame e o lood isk House ele a ion (Aliagha e al. 2015)Yes Expe ienced lood impac s (Hossain e al. 2022, Osbe ghaus and Rei 2021, Paopid e al. 2020, Senapa i 2020a, Liu e al. 2019, Fahad and Jing 2018, Reynaud e al. 2018, A shad e al. 2016, Oulahen 2015, A eya e al. 2015, Tu ne e al. 2014, Sei e e al. 2013, Hung 2009, B owne and Hoy 2000) Yes Flood dep h and du a ion (Paopid e al. 2020, Aliagha e al. 2015)Yes P esence o o he isk- educ ion measu es/le ee p o ec ion (Hossain e al. 2022; This le hwai e e al. 2020; Kousky 2011)Yes 6 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Table 1 (con inued) Ca ego y Thema ic a ea Pa ame e Re e ences Compa able pa ame e in su ey da a se In e ac ion In e ac ion wi h insu ance ins i u ions P ice o insu ance (Na ud and Vondolia 2020; Reynaud e al. 2018; B owne and Hoy 2000)No Mul i-yea insu ance policies/billing equency (Reynaud e al. 2018; Bo zen e al. 2013)No The amoun o e ed in he insu ance con ac (Senapa i 2020a; Reynaud e al. 2018)No T us in insu e s (Sidi e al. 2018; Reynaud e al. 2018; Aliagha e al. 2014)Yes Types o isk co e ed (Reynaud e al. 2018)Yes P e ious insu ance pu chase (Senapa i 2020a)Yes Insu ance p o ide (Reynaud e al. 2018)Yes Pe cep ion o e ec i eness o insu ance (Abbas e al. 2015)Yes Awa eness o insu ance (unde s anding) (Oduniyi e al. 2020; Senapa i 2020b)Yes In e ac ion wi h o he ins i u ions & social en i onmen Pe cei ed esponsibili y o p e en ing damage (Oulahen 2015)Yes Humani a ian/public compensa ion (Sei e e al. 2013, Bo zen and an den Be gh 2012)Yes Flood isk communica ion (Bo zen e al. 2013)Yes Flood p edic ion (wa ning) (Sidi e al. 2018)Yes Access o in o ma ion and ex ension se ices (Hossain e al. 2022; Adzawla e al. 2019)Yes Membe ship in a me ’s g oups (Hossain e al. 2022; Adzawla e al. 2019)Yes Pe cep ion owa ds go e nmen e o in handling lood (Sidi e al. 2018)Yes Risk sha ing be ween agen s (Be g e al. 2022)Yes Social in luence (Lo 2013)No 7 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Table 1 (con inued) Ca ego y Thema ic a ea Pa ame e Re e ences Compa able pa ame e in su ey da a se A ibu es A ibu es (o HH/indi idual) Income (Dewi e al. 2018, Sidi e al. 2018; A shad e al. 2016; Ren and Wang 2016; Aliagha e al. 2015, 2014; Abbas e al. 2015; Kousky 2011; Hung 2009; B owne and Hoy 2000) Yes Educa ion (Oduniyi e al. 2020; Adzawla e al. 2019; Sidi e al. 2018; A eya e al. 2015)Yes Age (Oduniyi e al. 2020; A eya e al. 2015; Abbas e al. 2015)Yes E hnici y (A eya e al. 2015)Yes A i udes owa ds isk aking (e.g., isk a e se) (Hossain e al. 2022; Reynaud e al. 2018, Bo zen and an den Be gh 2012)Yes In e nal locus o con ol (Robinson and Bo zen 2020)Yes Abili y o pay (Fahad and Jing 2018; A shad e al. 2016)Yes Al e na i e income sou ces (non-ag icul u al) (Hossain e al. 2022; Adzawla e al. 2019; Abbas e al. 2015)Yes P e e ence unce ain y (Hung 2009)Yes Conse a ism (Hung 2009)No Fa me ’s expe ience (Oduniyi e al. 2020)Yes Ma i al s a us (Oduniyi e al. 2020)Yes HH dependen s (Oduniyi e al. 2020)Yes Remi ances (Adzawla e al. 2019)Yes Ha ing he loca ion o he house in an a luen a ea (Adzawla e al. 2019)No 8 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Table 1 (con inued) Ca ego y Thema ic a ea Pa ame e Re e ences Compa able pa ame e in su ey da a se Po en ial asse s o be insu ed House p ice/dwelling alue (Da ling on and Yiannakoulias 2022, Paopid e al. 2020, Kousky 2011)No Amoun o land owned (Kousky 2011)Yes Land s a us (owne ship) (Dewi e al. 2018, Abbas e al. 2015)Yes Fa m ypology (Fahad and Jing 2018; A shad e al. 2016)Yes Cul i a ed land size (Senapa i 2020a)No Fa m size (Dewi e al. 2018)No Seed p ices (Senapa i 2020a)No Fe ilize p ices (Senapa i 2020a)No Expendi u e o a me (Dewi e al. 2018)No House condi ions (Hung 2009)Yes Comme cial p oduc ion (Adzawla e al. 2019)No 15 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Table 2 (con inued) Pa ame e s Responses F equency Pe cen age Addi ional sou ces o income (mul iple esponses possible) Raising ca le 213 28.6 Fishing 86 11.6 Hun ing 7 0.9 Local indus ies 188 25.3 Manu ac u ing indus ies 14 1.9 Cons uc ion and public wo ks 13 1.7 Comme ce, ca e ing and accomoda ion 182 24.5 T anspo and communica ion 26 3.5 Banks and insu ance 1 0.1 No esponse 91 12.2 Cu en ly owning any o m o insu ance Yes 17 2.3 No 727 97.7 To al 744 100 P e iously owned insu ance bu e mina ed he con ac Yes 8 1.1 No 736 98.9 To al 744 100 16 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Da a P epa a ion andVa iable Selec ion Ini ially, da a had o be sepa a ed in o ca ego ical and nume ical pa ame e s while clean- ing he da a and emo ing NaN (No a Numbe ) alues. The la e was necessa y since he p esence o NaN alues will s op he calcula ion o i ing he model i no emo ed, bu will also gene a e NaN alues a e calcula ion. Fo he c ea ion o he model, one-ho encoding was used o he ca ego ical pa ame e s ( ans o ma ion in o bina y 0–1 pa am- e e s) and s anda d scaling o he nume ical da a (disca ding mean and scaling acco ding o a iance o he uni ) o be able o c ea e a p ocesso o he model. The p ocess o pa ame e selec ion is illus a ed in Fig. 4. In o de o begin he ini- ial selec ion o ele an pa ame e s, ea u e columns we e assessed based on he p- alue and (Spea man) co ela ion alue o unco e he ela ionships be ween pa ame e s. This s eps allowed o a educ ion o he ini ially mo e han 400 pa ame e s o a ound 100. The emaining pa ame e s we e hen g ouped by opic in o he six a eas o he amewo k p e- sen ed in Fig.1. Then, pai plo s (showcasing pai wise bi a ia e dis ibu ions) and a (Pea - son) co ela ion hea map we e gene a ed o u he acili a e he selec ion o in luen ial pa ame e s. Based on he hea map co ela ion ma ix, i was decided o use he pa ame e s wi h low co ela ion alues while dis ega ding he o he s, as he high co ela ion pa am- e e s can be connec ed and ela ed in wo ways: i he alues o co ela ion a e highe han + 0.5, hen hese pa ame e s a e di ec ly co ela ed and i less han -0.5 hen hey a e in e sely co ela ed, which means i one pa ame e ends o inc ease, hen he connec ed one dec ease o nega i e alues while i inc eases o posi i e alues. Fo addi ional e i- ica ion, c oss- abula ions ha illus a e he co ela ions be ween he pa ame e s and he ou pu pa ame e we e used be o e u he s eps we e conduc ed in he analysis. Mo eo e , i allowed o deciding which pa ame e s o e ain o d op. Compa ison o Machine Lea ning Models Machine lea ning models we e es ed by using he Sciki -lea n sklea n package. Fo all models, he da a was spli in o aining (67%) and es da a (33%). The i s model was he mul inomial logis ic eg ession model, and is conside ed a supe ised lea ning ech- nique. This echnique se es o p edic i an objec belongs o a ce ain class by p o iding 111 285 50 33 17 11 96 34 90 17 0 50 100 150 200 250 300 Ve y likely Likely Indi e en Unlikely Ve y unlikely Togo (n=496)Benin (n=248) Fig. 3 Dis ibu ion o esponses wi hin ou come a iable (likelihood o pu chase o a po en ial lood insu - ance p oduc ) 17 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 a p obabili y on a ange be ween 0 and 1 (James e al. 2021). Fu he mo e, he His og am- based G adien Boos ing classi ie model was applied, which conside s g adien alues ob ained by p io upda e s eps om mo ing in o he s eepes di ec ion o descen (Feng e  al. 2018). Also hype pa ame e uning and g idsea ch we e applied o his classi ie , which howe e did no lead o a sa is ac o y imp o emen o he model accu acy. Finally, addi ional machine lea ning es s we e applied by using decision ees, a me hod d awing upon he Gini-Index (James e al. 2021). In addi ion, bagging was applied o he decision ees o lowe he a iance in he p edic ion unc ion, as well a andom o es model, d aw- ing upon an assembly o a ious decision ees (Has ie e al. 2009). Deep Lea ning Model (Sequen ial Neu al Ne wo k) In o de o a emp achie ing be e esul s han he ones ob ained om mo e con en- ional machine lea ning app oaches (see 3.2.2), his s udy added a deep lea ning (DL) model (sequen ial neu al ne wo k model) o he analysis using bo h he Tenso Flow and Ke as packages. Sequen ial models a e pa o a i icial neu al ne wo ks, which usu- ally consis o se e al laye s (inpu laye , hidden laye s, and ou pu laye ) ha each a e equipped wi h se e al nodes/neu ons, con aining ac i a ion unc ions, ha a e con- nec ed h ough weigh ed connec ions be ween he laye s (Jung 2022; James e al. 2021). In gene al, a sequen ial model p ocesses he inpu ed da a in a one-di ec ional, linea sequence om he inpu laye , passing h ough he hidden laye s, and a i ing a he ou - pu laye (Cholle 2021). Usually, DL app oaches a e chosen in cases whe e ex emely la ge da a se s a e p ocessed and when he possibili y o in e p e he model does no play and impo an ole (James e al. 2021). S ill, his s udy applied his app oach o cla i y i a DL model would imp o e he accu acy o p edic ion. Wi h ega ds o he la ge amoun o ca ego ical da a, ha we e encoded, i also helped o conside a la ge Fig. 4 Selec ion p ocess o he inal se o model pa ame e s 18 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 amoun o a ailable da a. To analyze nume ical and ca ego ical ea u es in a combined manne in his DL model, ea u e columns we e de ined by using a Dense Fea u es laye and using i as an inpu in o he Ke as model. The sequen ial model buil o his s udy uses he Relu (Rec i ied Linea Uni ) ac i a ion unc ion o he inpu laye , no allow- ing ac i a ion o he neu on i inpu alues a e below 0 (James e al. 2021), and a So - max unc ion o he ou pu laye , which is bes sui ed i a ca ego ical ou pu is desi ed (Klimo e al. 2021). Each neu on o he inpu laye ecei es a a iable o he da ase and passes ha in o ma ion o ano he neu on, which leads o a highe numbe o neu ons wi h a highe numbe o a iables. This model con ains 256 neu ons. Besides, he So - max laye mus ha e he same numbe o nodes as he ou pu laye , which is i e in he case o his model (Fig.5). The ac i a ion laye is ac ually he nonlinea unc ion and i ans o ms he alues o he i s hidden laye in o weigh ed sums o he nex laye . In addi ion, he Adam as op imize wi h a c oss en opy and 200 epochs was applied o i ing he model. To compa e his model, a second DL model was gene a ed con ain- ing 50 neu ons, he he_uni o m unc ion as ke nel ini ialize , d awing samples om a unca ed no mal dis ibu ion cen ed on 0 and he s ochas ic g adien descen (SGD) op imize . Sequen ial models bea he disad an age ha hey only allow o p o ide inpu in o he model only once a he beginning, in con as o unc ional models in which lay- e s can be connec ed o one ano he in a mul i-di ec ional way, allowing o eed-back loops (Cholle 2021). Ye , sequen ial models s ill be e allow o a conside a ion o a la ge numbe o inpu pa ame e s in compa ison o a con en ional eg ession model app oaches, as cu en ly widely used in he ield o WTP/WTI. In addi ion, in compa i- son o con en ional ML app oaches a neu al ne wo k can lea n om he da a in a be e and mo e complex way and e en wo k wi h uns uc u ed da a (Janiesch e  al. 2021) and hus be e e lec he esea ch con ex . This conside a ion was o high impo ance o his esea ch p ojec o no di ec ly in e indings and assump ions om s udies in egions wi h mo e es ablished insu ance ma ke s. Ins ead his s udy wan s o conside a wide ange o pa ame e s o be e ep esen he in e es le els o a popula ion ha has no been widely exposed o he usage o such p oduc s be o e. Fig. 5 Applica ion o So max on he DL model ou pu laye 19 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Resul s Selec ed Rele an Pa ame e s Acco ding oPai plo s, Co ela ion Ma ix andC oss Tabs Fo pa ame e selec ion, ea u e columns o he en i e da a se we e ini ially assessed o he en i e da a se . Then, he emaining pa ame e s we e ca ego ized in o he six hema ic a eas o he amewo k (Fig.1). Mo eo e , he g ouped pa ame e s we e assessed h ough pai plo s and a hea map co ela ion ma ix. As a inal s ep o e i ica ion, c oss abs we e used o assessing he co ela ion be ween he pa ame e s and he ou pu alue. The el- e an pa ame e s e lec ed all six hema ic a eas o he p esen ed amewo k on in luen ial ac o s on insu ance demand. As isualized in Table1, pa ame e s on po en ial asse s o be co e ed we e only spa sely ep esen ed in his da a se , which can be seen as he eason o hem only appea ing once in he inal selec ed se o pa ame e s. Finally, 38 pa ame e s (including one ou pu pa ame e ) make up he inal se o selec ed pa ame e s (Table3). The selec ed pa ame e s o he model co e ed he ollowing ca ego- ies o pa ame e s om he amewo k: Pe cep ion on clima e change; Flood isk pe cep- ion; Expe ienced lood impac s; (Ex e nally de ined) le el o lood isk; Awa eness o insu ance (unde s anding); T us in insu e s; Pe cep ion o e ec i eness o insu ance; P e- ious insu ance pu chase; Insu ance p o ide ; Types o isk co e ed; Pe cei ed esponsi- bili y o p e en ing damage; Humani a ian/public compensa ion; Membe ship in a me ’s g oups; Risk sha ing be ween agen s; Income; Ma i al s a us; Abili y o pay; P e e ence unce ain y; Land s a us (owne ship). Model Accu acies All models we e applied o h ee sepa a e da a se s each, namely one o e all da a se con- aining submissions om bo h Togo and Benin (n = 744) as well as wo subse s om Togo (n = 496) and Benin (n = 248) exclusi ely. Ini ially, six machine lea ning models we e un on he da a se s and compa ed by hei model accu acy. The applied model ypes o he classi ica ion a e logis ic eg ession, a his og am-based g adien boos ing classi ie , an op imized his og am-based g adien boos ing classi ie , decision ees, a bagging ees clas- si ie , and a andom o es classi ie . Mo eo e , a sequen ial neu al ne wo k was applied o he da a se s o compa e i a DL model would yield highe accu acies han he con en ional ML models. As illus a ed in Table4, almos all models (excep o he op imized his og am-based g adien boos ing classi ie ) e u ned he highes accu acies o he Togo subse . The logis- ic eg ession classi ie e u ned an accu acy o 54.0% (s d = 0.029) o he combined da a se , 48.0% (s d = 0.0042) o he Benin subse , and 61.7% (s d = 0.049) o he Togo sub- se . O e all, his classi ie he e o e anked among he ones wi h he weakes pe o mances o he con en ional ML models. The his og am-based g adien boos ing classi ie achie ed 64.0% (s d = 0.00) o he combined da a se , 55.5% (s d = 0.00) o he Benin subse , and 65.3% (s d = 0.00) o he Togo subse . Thus, i anked among he be e pe o m- ing con en ional ML models, especially o he combined da a se and he Benin subse . The model was e en imp o ed u he h ough hype pa ame e uning and applying g id sea ch. The model hen achie ed 67.0% (s d = 0.00) accu acy o he combined da a se , 58% pe cen (s d = 0.00) o he Benin subse , which we e he highes o all con en ional 20 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Table 3 Summa y o included model pa ame e s o assessing he demand o lood insu ance Ca ego y Thema ic a ea Associa ed ca ego y o pa ame e s om amewo k Desc ip ion o selec ed pa ame e s om he su ey da a se Flood isk “Subjec i e” pe cep ion o lood isk Pe cep ion on clima e change In e iewee hea d o clima e change be o e Flood isk pe cep ion Pe cei ed likelihood o u u e looding “Objec i e” Flood Risk Expe ienced lood impac s Financial eco e y ime om comme cial impac s F equency o comme cial impac s (pas 20yea s) In ensi y o comme cial impac s (pas 20yea s) Financial eco e y ime om all ou impac ypes combined F equency o all ou impac ypes combined (pas 20yea s) Se e i y o all ou impac ypes combined (pas 20yea s) (Ex e nally de ined) le el o lood isk Flood isk zone based on dis ance o he i e , ele a ion, and epo s o lood a ec edness In e ac ion In e ac ion wi h insu ance ins i u ions Awa eness o insu ance (unde s anding) Unde s anding o how insu ance wo ks No p e ious insu ance pu chase due o lack o in o ma ion T us in insu e s Le el o us ha insu ance companies will deli e payou as p omised No p e ious insu ance pu chase due o gene al lack o us in companies Pe cep ion o e ec i eness o insu ance Insu ance as an ins umen only sui ed o he needs o weal hy people No p e ious insu ance pu chase due o oo much pape wo k P e ious insu ance pu chase Household has access o insu ance in case o expe iencing lood impac s Insu ance p o ide No insu ance p o ide /p oduc s p esen in he a ea Types o isk co e ed Desi ed isk o be co e ed in po en ial lood insu ance p oduc In e ac ion wi h o he ins i u ions & social en i onmen Pe cei ed esponsibili y o p e en ing damage Desi ing o ha e access o emi ances o deal wi h lood impac s Humani a ian/public compensa ion Household has access o go e nmen al suppo in case o expe iencing lood impac s Household has access o NGO suppo in case o expe iencing lood impac s Membe ship in a me ’s g oups Household has access o suppo om coope a i es in case o expe iencing lood impac s Risk sha ing be ween agen s Household is membe o a sa ings g oup Household has access o c edi s om banks in case o expe iencing lood impac s Household d aws upon hei own esou ces in case o expe iencing lood impac s Household has access o suppo om communi y solida i y unds in case o expe iencing lood impac s Household has access o c edi s om sa ings g oups in case o expe iencing lood impac s Household has access o c edi s om a p i a e lende in case o expe iencing lood impac s Household has no access o any p e iously men ioned sou ce in case o expe iencing lood impac s Household has no bough any insu ance p e iously because hey had access o o he mechanisms o co e - age 21 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Table 3 (con inued) Ca ego y Thema ic a ea Associa ed ca ego y o pa ame e s om amewo k Desc ip ion o selec ed pa ame e s om he su ey da a se A ibu es A ibu es (o HH/indi idual) Income Household income pe yea Ma i al s a us Household is emale-headed Abili y o pay Fea ha insu ance pu chase will a ec mo e essen ial needs o he household o be co e ed Household has no bough any insu ance be o e due o lack o means P e e ence unce ain y Household has no bough any insu ance be o e due o no being in e es ed in he opic Unce ain y on he eason why no insu ance has been pu chased be o e Asse s o be co e ed Land s a us (owne ship) Household is owne o he house hey a e li ing in 22 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Table 4 Model accu acies o ML/DL models applied o he selec ed pa ame e s Con en ional Machine lea ning Deep lea ning Logis ic Reg es- sion His og am-based G adien Boos ing Classi ie Op imized His og am-based G adien Boos ing Classi ie Decision T ees Bagging ees classi ie Random Fo es Classi ie Sequen ial Neu al Ne - wo k Fi s model Sequen ial Neu al Ne wo k second model Accu acy bo h coun ies (n = 744) 0.540 ± 0.029 0.64 ± 0.000 0.67 ± 0.000 0.437 ± 0.034 0.612 ± 0.045 0.636 ± 0.035 1 ± 5.67 × 10–5 0.9350 ± 0.2329 Accu acy Benin subse (n = 248) 0.480 ± 0.0042 0.550 ± 0.000 0.58 ± 0.000 0.476 ± 0.051 0.552 ± 0.035 0.585 ± 0.048 1 ± 0.0013 0.9756 ± 0.1614 Accu acy Togo subse (n = 496) 0.617 ± 0.049 0.653 ± 0.000 0.69 ± 0.000 0.534 ± 0.049 0.704 ± 0.041 0.716 ± 0.051 1 ± 8.17 × 10–5 0.9512 ± 0.1291 23 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 ML models, and 69% (s d = 0.00) o he Togo subse . Mo eo e , a decision ee clas- si ie was applied, which me ely eached 43.7% (p = 0.034) o he combined da a se , 47.6% (s d = 0.051) o he Benin subse , and 53.4% (s d = 0.049) o he Togo subse . As a consequence, his classi ie achie ed he lowes accu acies among all con en ional ML models. Howe e , i was imp o ed by applying bagging o hen each 61.2% (s d = 0.043) o he combined da a se , 55.2% (s d = 0.035), and e en 70.4% (s d = 0.041) o he Togo subse . Finally, as he las con en ional ML model, a andom o es classi ie was applied achie ing 63.6% (s d = 0.035) o he combined da a se , 58.5% (s d = 0.048) o he Benin subse , and e en 71.6% (s d = 0.051) o he Togo subse . These esul s clea ly show ha he da ase s o Togo ende ed he highes accu acies. The la e is due o he ac ha he e is highe co ela ion in he answe s p o ided by esponden s in Togo. Since he accu acies o he con en ional ML models did no yield highe accu acies (o e 75–80%), wo sequen ial neu al ne wo ks om he ealm o DL we e applied as a compa ison. The i s sequen ial neu al ne wo k model e u ned 100.0% o accu acy o he combined da a se , as well as o he Benin and Togo subse s. As a consequence, i yielded he bes pe o mance by a in compa ison o he applied con en ional ML models. This inding eme ged somewha su p ising, since deep lea ning is a he ecommended o da a se s ha a e much la ge han he su ey da a se . The second model howe e exhibi ed a sligh ly lowe accu acy wi h 93.5% o he combined da a se , 97.6% o he Benin subse and 95.12% o he Togo subse . A mo e de ailed o e iew on he loss, p ecision, F1 sco e and ecall a e p o ided in Annex 1 as well as a con usion ma ix in Annex 2 in he supple- men a y in o ma ion o his a icle. Con ibu ion o Pa ame e s oP edic ing Likelihoods o Insu ance Pu chase in heDeep Lea ning Model Fo he sequen ial neu al ne wo k model an o e iew o he mos impo an pa ame e s based on he ea u e impo ance alue was gene a ed (Fig.6). The ea u e impo ance alue exp esses he le el o in luence o a pa ame e on he ou pu a iable o he model (likeli- hood o insu ance pu chase). When iden i ying he mos impo an ea u es, a subse o ele an ea u es can be selec ed o use in building a model. The e o e, he dimensionali y is educed as well as noise in he da a. Mo eo e , he model in e p e abili y is imp o ed in ha way. The selec ion o ea u e impo ance u he mo e assis s in educing he numbe o pa ame e s, he e o e educing he da a and dec easing he ime needed o ob ain he esul s. The ea u e impo ance alues we e gene a ed o he combined da a se o bo h coun ies, as well as o he Togo and Benin subse s. In gene al, i can be obse ed ha he ea u e impo ance a ies in pa s o a la ge ex en ac oss he pa ame e s o he indi idual da a se s. Wi h ega ds o he pa ame e ca ego ies ou lined by he amewo k p esen ed in he s udy, in e ac ion- ela ed pa ame e s we e he mos impo an ca ego y o pa ame e s by a . Impo an pa ame e s ela ed o he hema ic a ea o in e ac ion wi h insu ance ins i u- ions we e he desi ed isk (ag icul u al, ma e ial, heal h, o comme cial impac s) o be co e ed in po en ial lood insu ance p oduc (Togo). Also, he deg ee o which insu ance was pe cei ed as an ins umen only sui ed o he needs o weal hy people (all) exhibi ed a high ea u e impo ance. In addi ion, pa ame e s ela ing he in e ac ion wi h o he ins i- u ions and he social en i onmen eme ged as he hema ic a ea wi h he mos nume ous impo an alues. Fea u e impo ance was high when a household had no access o any sou ce men ioned in he ques ionnai e o inancial coping in case o expe iencing lood 24 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 070605040302010 Likelihood o u u e looding (pe cep ion) Clima e change knowledge Financial eco e y ime (all) F equency o impac s (all) Se e i y o impac s (all) Financial eco e y ime (comme cial) F equency o impac s (comme cial) In ensi y o impac s (comme cial) Flood isk le el Le el o comp ehension Lack o in o ma ion T us in se ice deli e y Lack o us in companies Pe cep ion o insu ance Expec ed pape wo k Access o insu ance Lack o p esence Risk ype co e ed by insu ance Remi ances (desi ed o coping) Access o go e nmen al suppo Access o NGO suppo Access o coope a i es Household sa ings g oup membe Access o c edi s (bank) Coping wi h own means Access o communi y solida i y unds Access o c edi s (sa ings g oups) Access o c edi s (p i a e lende ) No access o any sou ce Access o o he sou ces Income pe yea Female-headed household P io i izing mo e essen ial needs Lack o means Lack o in e es Unce ain y House owne ship “Subje c i e” pe cep ion o lood isk “Objec i e” Flood Risk In e ac ion wi h insu ance ins i u ions In e ac ion wi h o he ins i u ions & social en i onmen A ibu es (o HH/indi idual) Ass e s o be co e e d Flood iskIn e ac ionA ibu es Fea u e impo ance all Fea u e impo ance Togo Fea u e impo ance Benin Fig. 6 Fea u e impo ance o pa ame e s in he sequen ial neu al ne wo k model 31 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Oulahen G (2015) Flood insu ance in Canada: implica ions o lood managemen and esiden ial ulne - abili y o lood haza ds. En i on Manage 55:603–615. h ps:// doi. o g/ 10. 1007/ s00267- 014- 0416-6 Paopid S, Tang J, Leelawa N (eds) (2020) Willingness o pay o lood insu ance: a case s udy in Phang Khon, Sakon Nakhon P o ince, Thailand. IOP Con e ence Se ies: Ea h and En i onmen al Science Pa koo EN, Thiam S, Adjonou K, Kokou K, Ve leysdonk S, Adounkpe JG, Villamo GB (2022) Com- pa ing expe and local communi y pe spec i es on lood managemen in he lowe Mono Ri e Ca chmen , Togo and Benin. Wa e 14:1536. h ps:// doi. o g/ 10. 3390/ w1410 1536 Pill M (2022) Towa ds a unding mechanism o loss and damage om clima e change impac s. Clim Risk Manag 35:100391. h ps:// doi. o g/ 10. 1016/j. c m. 2021. 100391 Ranasinghe R, Ruane AC, Vau a d R, A nell N, Coppola E, C uz FA, Dessai S, Islam AS, Rahimi M, Ruiz D (2021) Clima e Change In o ma ion o Regional Impac and o Risk Assessmen . In: Mas- son-Delmo e V, Zhai P, Pi ani A, Conno s SL, Péan C, Be ge S, Caud N, Chen Y, Gold a b L, Gomis MI, Huang M, Lei zell K, Lonnoy E, Ma hews J, Maycock TK, Wa e ield T, Yelekçi O, Yu R, Zhou B (eds) Clima e Change 2021: The Physical Science Basis. Con ibu ion o Wo king G oup I o he Six h Assessmen Repo o he In e go e nmen al Panel on Clima e Change. Cam- b idge Uni e si y P ess, Camb idge Ren J, Wang HH (2016) Ru al homeowne s’ willingness o buy lood insu ance. Eme g Ma k Financ T ade 52:1156–1166. h ps:// doi. o g/ 10. 1080/ 15404 96X. 2015. 11348 67 Reynaud A, Nguyen M-H, Aube C (2018) Is he e a demand o lood insu ance in Vie nam? Resul s om a choice expe imen . En i on Econ Policy S ud 20:593–617. h ps:// doi. o g/ 10. 1007/ s10018- 017- 0207-4 Robinson PJ, Bo zen WJW (2019) De e minan s o p obabili y neglec and isk a i udes o disas e isk: an online expe imen al s udy o lood insu ance demand among homeowne s. Risk Anal 39:2514–2527. h ps:// doi. o g/ 10. 1111/ isa. 13361 Robinson PJ, Bo zen W (2020) Flood insu ance demand and p obabili y weigh ing: he in luences o eg e , wo y, locus o con ol and he h eshold o conce n heu is ic. Wa e Resou Econ 30:100144. h ps:// doi. o g/ 10. 1016/j. w e. 2019. 100144 Schä e L, Wa ne K, K e S (2019) Explo ing and managing adap a ion on ie s wi h clima e isk insu - ance. In: Mechle R, Bouwe LM, Schinko T, Su minski S, Linne oo h-Baye J (eds) Loss and damage om clima e change. Sp inge In e na ional Publishing, Cham, pp 317–341 Sei e I, Bo zen WJW, K eibich H, Ae s JCJH (2013) In luence o lood isk cha ac e is ics on lood insu ance demand: a compa ison be ween Ge many and he Ne he lands. Na Haza ds Ea h Sys Sci 13:1691–1705. h ps:// doi. o g/ 10. 5194/ nhess- 13- 1691- 2013 Senapa i AK (2020b) Insu ing agains clima ic shocks: E idence on a m households’ willingness o pay o ain all insu ance p oduc in u al India. In J Disas e Risk Reduc 42:101351. h ps:// doi. o g/ 10. 1016/j. ijd . 2019. 101351 Senapa i AK (2020a) Do a me s alue insu ance agains ex eme d ough s and loods? E idence om Odi- sha, India. Glob Bus Re : 097215092095761. h ps:// doi. o g/ 10. 1177/ 09721 50920 957616 Sidi P, Mama MB, Sukono, Supian S, Pu a AS (2018) Demand analysis o lood insu ance by using logis- ic eg ession model and gene ic algo i hm. IOP Con Se Ma e Sci Eng 332:12053. h ps:// doi. o g/ 10. 1088/ 1757- 899X/ 332/1/ 012053 This le hwai e J, Hens a D, B own C, Sco D (2020) Ba ie s o insu ance as a lood isk managemen ool: e idence om a su ey o p ope y owne s. In J Disas e Risk Sci 11:263–273. h ps:// doi. o g/ 10. 1007/ s13753- 020- 00272-z T isos CH, Adelekan IO, To in E, Ayanlade A, E i e J, Gemeda A, Kalaba K, Lenna d C, Masao C, Mgaya Y (2022) A ica. In: Pö ne H-O, Robe s DC, Tigno M, Poloczanska ES, Min enbeck K, Aleg ía A, C aig M, Langsdo S, Löschke S, Mölle V, Okem A, Rama B (eds) Clima e Change 2022: Impac s, Adap a ion and Vulne abili y. Con ibu ion o Wo king G oup II o he Six h Assessmen Repo o he In e go e nmen al Panel on Clima e Change. Camb idge Uni e si y P ess, Camb idge, pp 1285–1455 Tu ne G, Said F, A zal U (2014) Mic oinsu ance demand a e a a e lood e en : e idence om a ield expe imen in Pakis an. Gene a Pap Risk Insu Issues P ac 39:201–223. h ps:// doi. o g/ 10. 1057/ gpp. 2014.8 Wagne S, Sou igne M, Walz Y, Balogun K, Komi K, K e S, Rhyne J (2021) When does isk become esidual? A sys ema ic e iew o esea ch on lood isk managemen in Wes A ica. Reg En i on Change 21:84. h ps:// doi. o g/ 10. 1007/ s10113- 021- 01826-7 Wagne S, Thiam S, Dossoumou NIP, Hagenloche M, Sou igne M, Rhyne J (2022) Reco e ing om inancial implica ions o lood impac s— he ole o isk ans e in he Wes A ican con ex . Sus ain- abili y 14:8433. h ps:// doi. o g/ 10. 3390/ su141 48433 Wanyan R, Yang L, Pu M, Zhao T, Zeng L (2022) The nexus be ween ai pollu ion and li e insu ance demand in china: e idence om deep machine lea ning. In: Sun X, Zhang X, Xia Z, Be ino E (eds) 32 Economics o Disas e s and Clima e Change (2024) 8:1–32 1 3 Ad ances in a i icial in elligence and secu i y, ol 1586. Sp inge In e na ional Publishing, Cham, pp 524–539 Publishe ’s No e Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in published maps and ins i u ional a ilia ions. Au ho s and A ilia ions SimonWagne 1,2· SophieThiam3· NadègeI.P.Dossoumou4· Da idDaou2 * Simon Wagne s6siw[email p o ec ed] 1 Ag icul u al Facul y, Uni e si y o Bonn, Meckenheime Allee 174, 53115Bonn, Ge many 2 Uni ed Na ions Uni e si y – Ins i u e o En i onmen andHuman Secu i y (UNU-EHS), UN Campus Pla z de Ve ein en Na ionen 1, D-53113Bonn, Ge many 3 Cen e o De elopmen Resea ch (ZEF) - Zen um ü En wicklungs o schung, Gensche allee 3, 53113Bonn, Ge many 4 Wes A ican Science Se ice Cen e onClima e Change andAdap ed Land Use (WASCAL) B.P., 1515Lomé, Togo