When econ­o­mists look to the sky

Econ­o­mists are in­creas­ingly turn­ing to data from satel­lite images to es­ti­mate the level of eco­nomic ac­tiv­ity in re­gions where eco­nomic data is er­ratic or un­avail­able

Mint Asia ST - - News - BSY UMIT M ISHRA

One key in­no­va­tion by the fi­nance min­istry in its an­nual Eco­nomic Sur­vey has caught ev­ery­one’s attention this year: the use of satel­lite data. The first vol­ume of the Sur­vey re­leased in Fe­bru­ary had used data from satel­lite images to cal­cu­late built-up area and es­ti­mate po­ten­tial prop­erty tax col­lec­tions to show how Indian cities are fail­ing to t ap a lu­cra­tive source of rev­enue (bit.ly/2x1o7cf). The sec­ond vol­ume re­leased in Au­gust uses satel­lite data to show that In­dia may be more ur­ban­ized than thought pre­vi­ously ( bit.ly/2x39q9k).

The use of satel­lite data is new in Indian pol­i­cy­mak­ing but a grow­ing body of eco­nomic re­search has re­lied on min­ing such data over the past few years to an­swer ques­tions re­lat­ing to growth and poverty in re­gions where of­fi­cial data is ei­ther un­avail­able or un­re­li­able. What started as a satel­lite pro­gramme run by the US depart­ment of defence to gauge cloud cover in the 1960s has in­creas­ingly be­come an im­por­tant re­source for econ­o­mists.

Night-time lights or night-lights data con­tain the data on en­ergy emit­ted or re­flected back from the sur­face of the earth to the sky. Econ­o­mists fig­ured that this data tends to cor­re­late with eco­nomic ac­tiv­ity. Thus, for coun­tries with poor data, night-lights data has been par­tic­u­larly use­ful in es­ti­mat­ing the level of eco­nomic ac­tiv­ity and gross do­mes­tic prod­uct (GDP). A clas­sic ex­am­ple is Myan­mar, which stopped pub­lish­ing its na­tional ac­counts statis­tics in 1989. Us­ing night-lights data, econ­o­mists have shown that the coun­try’s econ­omy is grow­ing at a very slow pace. A 2012 re­search pa­per (bit.ly/2gb9qe4) by the Ja­pan-based think-tank IDE-JETRO used night-lights data to show that most eco­nomic ac­tiv­ity is con­cen­trated in re­gions sur­round­ing the cap­i­tal Yan­gon. Also, the re­gions bor­der­ing China and Thai­land grew at much faster pace than those bor­der­ing Bangladesh and In­dia, the study sug­gests.

There has been sim­i­lar un­cer­tainty about growth in North Korea af­ter sanc­tions were im­posed against it. Night-lights data have once again come to the res­cue. As a cel­e­brated im­age (go.nasa.gov/ 1c545wo) con­firms, there is a stark dif­fer­ence be­tween South and North Korea; the North is al­most com­pletely dark while the South seems to be well-lit. A re­cent study (stan­ford.io/2vck­suv) by Yong Suk Lee of Stan­ford Uni­ver­sity us­ing night-lights data shows that eco­nomic ac­tiv­ity ap­pears to be con­cen­trated in ur­ban ar­eas, and par­tic­u­larly so in the cap­i­tal Py­ongyang. Lee found that eco­nomic sanc­tions de­creased lu­mi­nos­ity in the hin­ter­lands but in­creased lu­mi­nos­ity in ur­ban ar­eas, es­pe­cially the ur­ban core, sug­gest­ing that the dic­ta­tor­ship may have “coun­tered the ef­fects of sanc­tions by re­al­lo­cat­ing re­sources to the ur­ban ar­eas”. Also, eco­nomic ac­tiv­ity as mea­sured by lu­mi­nos­ity or night-lights seemed to have in­creased in re­gions bor­der­ing China, Lee’s study showed.

In In­dia, night-lights data has been used to un­der­stand the ef­fects of the re­or­ga­ni­za­tion of states. In 2000, three new states were cre­ated—ut­tarak­hand, Jhark­hand and Ch­hat­tis­garh. A 2015 study (bit.ly/2wv97as) by Sam Asher of the World Bank and Paul Novosad of Dart­mouth Col­lege, us­ing cen­sus and night­lights data, sug­gested that there has been marked im­prove­ment in eco­nomic ac­tiv­ity in the newly cre­ated states.

A re­cent re­search pa­per by (bit.ly/2vcol7l) an­a­lysts Praveen Chakravarty and Vivek De­he­jia of the IDFC In­sti­tute in Mum­bai used night-lights data to show that both in­ter-state and in­tra-state in­equal­ity in In­dia has been grow­ing.

The Eco­nomic Sur­vey this year also doc­u­mented the widen­ing re­gional in­equal­ity in In­dia, and at­trib­uted it to the dif­fer­ing qual­ity of gov­er­nance. But Chakravarty and De­he­jia ar­gue that much of the dif­fer­ences in eco­nomic ac­tiv­ity may be driven by network ef­fects.

“The con­ven­tional chan­nel of eco­nomic con­ver­gence is di­min­ish­ing re­turns to cap­i­tal, but we would ar­gue that this is off­set by the op­po­site phe­nom­e­non, of ag­glom­er­a­tion economies in cap­i­tal ac­cu­mu­la­tion and network ex­ter­nal­i­ties which in­crease, rather than de­crease, the mar­ginal pro­duc­tiv­ity of cap­i­tal as its stock in­creases,” the duo write. “…That is why Ap­ple has cho­sen to lo­cate its new man­u­fac­tur­ing unit in In­dia in wealthy and ex­pen­sive Kar­nataka rather than (say) poor and cheaper Bi­har, and why, by ex­ten­sion, it has cho­sen to lo­cate in wealthy and ex­pen­sive Ben­galuru rather than (say) poor and cheaper Shi­moga. If low labour costs and a pu­ta­tively higher re­turn to cap­i­tal were driv­ers, as con­ven­tional the­ory sug­gests, we ought to have seen Ap­ple lo­cate in Bi­har rather than Kar­nataka or at least within Shi­moga rather than Ben­galuru.”

There are two se­ri­ous lim­i­ta­tions of night­lights data that re­searchers have been grap­pling with for some time. One, satel­lites that record this data do not have the ca­pac­ity to de­tect ar­ti­fi­cial lights with pre­ci­sion and only cap­ture the lights emit­ted from ve­hic­u­lar traf­fic, rooftops and streets. The sec­ond is more se­ri­ous: night-lights data do not dis­tin­guish be­tween the poor­est and the poor re­gion. Be­yond a thresh­old, all is dark in the satel­lite images.

Re­search by Char­lotta Mel­lan­der of the Jönköping In­ter­na­tional Busi­ness School (bit.ly/2gbguf3) and co-au­thors, based on a study in Swe­den, sug­gests that while the cor­re­la­tion be­tween night-time lights and eco­nomic ac­tiv­ity is strong enough to make it a rel­a­tively good proxy for pop­u­la­tion and es­tab­lish­ment den­sity, the cor­re­la­tion is weaker in re­la­tion to wages. The re­searchers found the link be­tween light and eco­nomic ac­tiv­ity, es­pe­cially es­ti­mated by wages, to be “slightly over­es­ti­mated in large ur­ban ar­eas, and un­der­es­ti­mated in ru­ral ar­eas”.

To get around some of these is­sues, re­searchers have be­gun to use data from day­time satel­lite images, which are fed into a ma­chine learn­ing al­go­rithm to es­ti­mate lev­els of poverty, cap­i­tal stock, and eco­nomic ac­tiv­ity (bit.ly/2gu8qhw). Mul­ti­ple images, cap­tured on sev­eral dif­fer­ent days, are of­ten com-


Re­gional in­equal­ity: Night-time views of In­dia and sur­round­ing ar­eas in 2012 (top) and 2016. Econ­o­mists be­lieve that night-time lights or night-lights data tends to cor­re­late with eco­nomic ac­tiv­ity.

Newspapers in English

Newspapers from Malaysia

© PressReader. All rights reserved.