Subscribe to email updates...

Links

Marcus on wikipedia
Marcus on Facebook
Marcus on Smashwords
Marcus on NOOK Books
Marcus on National Library

Rain Predicts GDP

posted by Marcus, November 2, 2019 @ 5:26 am

Fore­cast­ing the cost of cli­mate change is a big deal, for obvi­ous rea­sons. But the eco­nom­ic mod­els soon turn unwieldy. Health impacts, fires, floods, drought, mud slides, salt water in ground­wa­ter, crop loss, infra­struc­ture costs, labour pro­duc­tiv­i­ty, migra­tion, ocean acid­i­fi­ca­tion, less food, more peo­ple, eco­nom­ic out­put, ener­gy con­sump­tion… The impacts are end­less and inter­re­lat­ed, fold­ing back on them­selves. For exam­ple, if it’s warmer pow­er plants need more water to cool them­selves, to gen­er­ate more pow­er to use in their own cool­ing and… Brain explo­sion. The chal­lenge with the mod­els is quan­ti­fy­ing mul­ti-var­ied impacts on every indus­try, plus geopo­lit­i­cal impacts like immi­gra­tion and con­flict, the cost of decar­bon­i­sa­tion, while fore­cast­ing pop­u­la­tion growth linked to health impacts and eco­nom­ic out­put and… you get the idea.

In 2006 the Stern Review had a run at it, and came up with a loss of 5%-but-maybe-20%-of-world-GDP. Stern said we could spend 1% of GDP if we act­ed fast and save 5%-but-maybe-20% in loss­es but that was 13 years ago so the num­bers are now out of date. Along comes Nord­haus, punch­es some patchy old inputs that came from Richard Tol into his mod­el, wins a Nobel Prize and says cli­mate change will cost $134 tril­lion by 2100 or we can spend $20 tril­lion now to save $30 tril­lion lat­er, then admits his mod­el is prob­a­bly bro­ken in his user man­u­al. Mean­while, Burke and his pals have a brain­wave. Instead of try­ing to fig­ure out all the thou­sands of ways glob­al warm­ing will cost mon­ey, let’s just see if hot coun­tries are poor­er than cold coun­tries. The UN likes Burke’s style and uses his method to fore­cast cli­mate change will cost $33 tril­lion by 2050, or 10% of glob­al GDP but we act fast we can save $12 tril­lion.

It’s here, read­ing Burke’s mod­el, that some­thing feels miss­ing. Cli­mate change has big, non-lin­ear that tran­scend polit­i­cal maps. One of the largest is that 16 bil­lion peo­ple in south east Asia rely on riv­er flows fed by declin­ing glacial melt water in the Himalayas. If the tem­per­a­ture changes in Nepal then Chi­na and India suf­fer. With these two coun­tries set to be the largest economies in the world by mid-cen­tu­ry, a sub­se­quent water deficit in these two nations will also have a mate­r­i­al impact on glob­al trade.

So how do you link the sig­nif­i­cance of water avail­abil­i­ty, rather than tem­per­a­ture, back to GDP? Attempt­ing to answer this ques­tion, I found a 2006 paper by Casey Brown and Upmanu Lall titled ‘Water and eco­nom­ic devel­op­ment: The role of vari­abil­i­ty and a frame­work for resilience’.

Fig­ure from Brown & Lall 2006

What is strik­ing about this paper is the cor­re­la­tion (shown abovE) between GDP per capi­ta and rain­fall vari­abil­i­ty, or ‘CVM’. Not total pre­cip­i­ta­tion. Vari­abil­i­ty.

So I set out to repro­duce Brown and Lall and acci­den­tal­ly made an unusu­al dis­cov­ery. First, if you exclude resource rents from GDP (using World Bank data on coal, oil and gas exports) the cor­re­la­tion strength­ens. These exports and in par­tic­u­lar oil, skew GDP per capi­ta sig­nif­i­cant­ly. Equal­ly, if the goal is to mod­el long-term sen­si­tiv­i­ty of economies to some­thing like weath­er, fos­sil fuel exports add addi­tion­al uncer­tain­ty to future decar­bon­i­sa­tion sce­nar­ios. Then, where Brown and Lall had used month­ly rain­fall vari­abil­i­ty, I tried out WRI Aque­duct Water Risk Atlas 3‑monthly, or sea­son­al, vari­abil­i­ty. The results are sim­i­lar, but the out­liers on the top-left side of the fig­ure Brown/Lall fig­ure above are now gone, and the role of total pre­cip­i­ta­tion as a fac­tor is even fur­ther dimin­ished.

Analy­sis of rain­fall vari­abil­i­ty vs GDP per capi­ta

The strik­ing thing about this result is that the rain­fall vari­abil­i­ty isn’t about volatil­i­ty or unpre­dictabil­i­ty, as 3‑monthly sea­son rain­fall vari­abil­i­ty repeats every year. The cor­re­la­tion seems inex­plic­a­ble. And there aren’t very many stud­ies look­ing at the impact of rain­fall vari­abil­i­ty on eco­nom­ic activ­i­ty. Lazo et al. 2011 found US GDP can move 3.4% in a year based on weath­er, but that lim­its eco­nom­ic impact to the nor­mal range of con­di­tions in North Amer­i­ca. We can also exclude some ini­tial con­clu­sions about eco­nom­ic impacts of vari­able rain­fall. It’s not about flood dam­age. Swap­ping out rain­fall vari­abil­i­ty for river­ine flood risk rat­ing by coun­try and the cor­re­la­tion decays:

River­ine flood risk vs GDP per capi­ta

Anoth­er pos­si­bil­i­ty is that rain­fall vari­abil­i­ty just hap­pens to cor­re­late with oth­ers fac­tor that dri­ve GDP. Like, rain­fall might be more vari­able in places where it’s cold­er. Or do these coun­tries hap­pen to be rich­er for some oth­er rea­son? The prob­lem with these the­o­ries is that while total pre­cip­i­ta­tion varies fair­ly con­sis­tent­ly by lat­i­tude, rain­fall vari­abil­i­ty is based on a range of geo­graph­ic fea­tures like topog­ra­phy.

Rain­fall vari­abil­i­ty map from Her­rero et al 2016

Economies also derive water from rivers, glacial melt water and aquifers. But as with total pre­cip­i­ta­tion, the cor­re­la­tion with GDPs falls apart.

So what could it be about sea­son­al rain­fall vari­abil­i­ty that bears so strong­ly on wealth?

Eco­nom­ic Impacts of Vari­able Rain­fall

There are some obvi­ous eco­nom­ic impacts of high­ly vari­able rain­fall, and some not so obvi­ous. Obvi­ous­ly drought, flood­ing and extreme storms all incur costs. But by them­selves they aren’t enough to explain the glob­al dif­fer­ences in wealth. So too vari­able rain­fall affects agri­cul­ture – espe­cial­ly graz­ing. Spa­tiotem­po­ral rain­fall vari­abil­i­ty also impacts tourism at a macro­eco­nom­ic scale, and recre­ation­al and sport­ing events at a micro-eco­nom­ic lev­el. Min­ing and con­struc­tion – which both also con­sume water in their oper­a­tions – stop oper­at­ing com­plete­ly in heavy rain, incur­ring sig­nif­i­cant costs. So too, port and air­port oper­a­tions and fish­ing. All these impacts com­bined go some way to answer­ing the mys­tery. As for the less obvi­ous:

·        Infra­struc­ture: In an econ­o­my with high­ly vari­able pre­cip­i­ta­tion catch­ment and dis­tri­b­u­tion sys­tems, trans­port, ener­gy and telecom­mu­ni­ca­tions sys­tems all have to be engi­neered for a wider dis­tri­b­u­tion of con­di­tions, adding to design, con­struc­tion, main­te­nance and oper­at­ing costs. And not just water infra­struc­ture. In par­tic­u­lar, roads – both sealed and unsealed – are high­ly sus­cep­ti­ble to fluc­tu­a­tions in sub­soil and sur­face mois­ture. Rain also absorbs and scat­ters radio and radar sig­nals. In some economies addi­tion­al weath­er-proof­ing must be installed on radio tow­ers dur­ing mon­soon sea­son. This may also lead to an indi­rect impact on pro­duc­tiv­i­ty to impacts on air and sea nav­i­ga­tion.

·        Trans­port: In addi­tion to the dete­ri­o­rat­ing effects on pave­ments, rain is the num­ber one cause of road acci­dentscon­firmed by mul­ti­ple stud­ies, e.g.: here, here and here. Rain caus­es road acci­dents through loss of vis­i­bil­i­ty, tyre fric­tion, spray from oth­er vehi­cles, and cog­ni­tive strain. Vari­able rain­fall like­ly also has an indi­rect impact on pro­duc­tiv­i­ty of the trans­port indus­try.

·         Prop­er­ty Dam­age: in addi­tion to vehi­cle acci­dents, con­vec­tive storms also cause hail dam­age – which is dis­pro­por­tion­ate­ly expen­sive com­pared to oth­er weath­er events. In the US, total dam­ages from con­vec­tive storms is the same as all hur­ri­canes.

·        Human Con­flict: Numer­ous stud­ies show the links between water avail­abil­i­ty and civ­il unrestciv­il warprotests and riots, and polit­i­cal repres­sion. But a 2015 study by sci­en­tists at Stan­ford, Berke­ley and Cam­bridge also found a cor­re­la­tion between indi­vid­ual mur­ders and rain­fall devi­a­tion. Sale­hyan & Hen­drix 2014, Von Uexkull 2014, Ghimire & Fer­reira 2016, Mares & Mof­fet­ti 2016, and Burke et al. 2009 all find sim­i­lar results on inter­per­son­al con­flict and vio­lent crime.

·        Oth­er inter­ac­tions: Extremes and deficits in rain­fall con­tribute to oth­er nat­ur­al dis­as­ters not count­ed in flood dam­age, like fire sea­son dura­tion and sever­i­ty. Soil ero­sion and drought also has an impact on land degra­da­tion with costs 17% of world GDP.

In addi­tion to all of the above, there is an addi­tion­al neg­a­tive feed­back in that many of the coun­tries most impact­ed by weath­er, with the low­est GDP also suf­fer a dis­pro­por­tion­ate lack of mete­o­ro­log­i­cal ser­vices fur­ther com­pound­ing the uncer­tain­ty in their hydro­log­i­cal cycles.

GDP per capi­ta (cir­cle size) vs rain­fall vari­abil­i­ty and total pre­cip­i­ta­tion. Top 50 economies in yel­low. Bot­tom 50 economies in red.

If rain­fall vari­abil­i­ty is such a strong pre­dic­tor of wealth, this has some seri­ous impli­ca­tions for fore­cast­ing the impacts of cli­mate change. But there is also a more philo­soph­i­cal les­son, if wealth is not entire­ly a prod­uct of polit­i­cal, social or behav­iour­al dif­fer­ence between nations. How much of our liv­ing con­di­tions are geo­graph­ic for­tune?

It is also with good rea­son that the world spends $50bn every year on know­ing where and when it will rain.

Leave a Reply