Decomposition analysis of earnings inequality in rural India: 2004-2012

Khanna, Shan anu; Goel, Deep i; Mo isse e, René A icle Decomposi ion analysis o ea nings inequali y in u al India: 2004-2012 IZA Jou nal o Labo & De elopmen P o ided in Coope a ion wi h: IZA Ne wo k @ LISER, Luxembou g Ins i u e o Socio-Economic Resea ch (LISER) Sugges ed Ci a ion: Khanna, Shan anu; Goel, Deep i; Mo isse e, René (2016) : Decomposi ion analysis o ea nings inequali y in u al India: 2004-2012, IZA Jou nal o Labo & De elopmen , ISSN 2193-9020, Sp inge , Heidelbe g, Vol. 5, Iss. 18, pp. 1-26, h ps://doi.o g/10.1186/s40175-016-0064-8 This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/169301 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/ ORIGINAL ARTICLE Open Access Decomposi ion analysis o ea nings inequali y in u al India: 2004–2012 Shan anu Khanna 1 , Deep i Goel 2,3* and René Mo isse e 4 * Co espondence: [email p o ec ed] 2 Depa men o Economics, Delhi School o Economics, Uni e si y o Delhi, 110007 Delhi, India 3 IZA-Ins i u e o Labo , Bonn, Ge many Full lis o au ho in o ma ion is a ailable a he end o he a icle Abs ac We analyze he changes in ea nings o paid wo ke s (wage ea ne s) in u al India om 2004/05 o 2011/12. Real ea nings inc eased a all pe cen iles, and he pe cen age inc ease was la ge a he lowe end. Consequen ly, ea nings inequali y declined. Recen e ed in luence unc ion decomposi ions show ha h oughou he ea nings dis ibu ion, excep a he e y op, bo h changes in “wo ke cha ac e is ics”and in “ e u ns o hese cha ac e is ics”inc eased ea nings, wi h he la e ha ing played a bigge ole. Decomposi ions o inequali y measu es e eal ha al hough he change in cha ac e is ics had an inequali y-inc easing e ec , chie ly a ibu able o inc eased educa ion le els, inequali y declined because wo ke s a lowe quan iles expe ienced g ea e imp o emen s in e u ns o hei cha ac e is ics han hose a he op. JEL: JEL Classi ica ion: J30, J31, O53 Keywo ds: Ea nings, Inequali y, Ea ning dis ibu ion, Ru al India 1 In oduc ion In hei discussion o India’s economic g ow h, Ko wal e al. (2011) poin o he exis - ence o wo Indias: “One o educa ed manage s and enginee s who ha e been able o ake ad an age o he oppo uni ies made a ailable h ough globaliza ion and he o he —a huge mass o unde educa ed people who a e making a li ing in low p oduc - i i y jobs in he in o mal sec o — he la ges o which is s ill ag icul u e.”This pape is abou he second India ha mainly esides in i s u al pa s. Ag icul u e, he mains ay o he u al economy, con inues o employ he la ges sha e o he Indian wo k o ce, bu i s con ibu ion o g oss alue added (GVA) is much smalle . In 2011, he employ- men sha es o ag icul u e, indus y, and se ices we e 49, 24 and 27 %, espec i ely, whe eas hei sha es in GVA we e 19, 33, and 48 %, espec i ely (GOI 2015). In addi ion, be ween 2004/05 and 2011/12, eal g oss domes ic p oduc (GDP) in hese sec o s g ew a 4.2, 8.5 and 9.6 % pe annum, espec i ely, making ag icul u e he slowes g owing sec o o he economy (au ho s’calcula ions based on RBI 2015). Gi en hese igu es, he conce n abou whe he high o e all GDP g ow h has bene i - ed hose a he bo om, and o wha ex en hey ha e bene i ed compa ed o hose a he op, is e en mo e pe inen o u al India. We he e o e ocus on u al India and examine how eal ea nings o paid wo ke s (wage ea ne s) e ol ed o e he 7-yea pe iod be ween 2004/05 and 2011/12. © The Au ho (s). 2016 Open Access This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0 In e na ional License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons license, and indica e i changes we e made. Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 DOI 10.1186/s40175-016-0064-8 Se e al s udies ha e documen ed ha along wi h he high g ow h a es o GDP ha ha e cha ac e ized he Indian economy since he 1980s, he e has been an inc ease in inequal- i y. 1 Howe e , mos o hese s udies ha e ei he ocused on consump ion expendi u e (Sen and Himanshu 2004; Cain e al. 2010; Mo i am and Vakulabha anam 2012; Jaya aj and Sub amanian 2015; Da e al. 2016) 2 o on ea nings o paid wo ke s in u ban India (Kijima 2006; Azam 2012a). Two no able excep ions a e Hna ko ska and Lahi i (2013) and Jacoby and Dasgup a (2015). Hna ko ska and Lahi i (2013) ocus on wage compa isons be ween u al and u ban a eas be ween 1983 and 2010. They ind ha u ban agglome a ion led o a massi e inc ease in u ban labo supply ha in u n educed he u al-u ban wage gap. Un- like Hna ko ska and Lahi i (2013), we ocus exclusi ely on u al India o p o ide a mo e de ailed pic u e o he changes wi hin his sec o . Jacoby and Dasgup a (2015) adop he supply-demand-ins i u ions (SDI) amewo k pionee ed by Ka z and Mu phy (1992) and Bound and Johnson (1992), o decompose wage changes be ween 1993 and 2011 in bo h u al and u ban India. We use a e y di e en app oach, namely, he ecen e ed in luence unc ion (RIF) decomposi ion de eloped by Fi po, Fo in, and Lemieux (2009) o s udy ea nings e olu ion in u al India. 3 Jacoby and Dasgup a (2015) decompose he change in an indi ec measu e o wage inequali y, namely, he ela i e wages o educa ed and unedu- ca ed wo ke s, in o changes in employmen sha es o di e en demog aphic g oups and changes in he indus ial composi ion. In his pape , we ocus on di ec measu es o in- equali y such as he Gini and he 90/10 pe cen ile a io, and decompose changes in hese measu es in o changes in wo ke cha ac e is ics and changes in e u ns o hese cha ac e - is ics. Ou inding ha he change in e u ns o cha ac e is ics is d i ing he decline in ea nings inequali y in u al India is a no el one. Mo eo e , we documen changes no jus a he mean bu also a a ious quan iles. I is impo an o do so because se e al s udies ha e ound ha ea nings inequali y is mainly concen a ed a he uppe end. Fo India, Azam (2012a) and Kijima (2006) ind his o u ban wage ea ne s and Bane jee and Pike y (2005) ind i o income ax paye s. We use uncondi ional quan ile eg essions o accoun o he e ec s o wo ke s’cha ac e is ics a di e en quan iles and he eby make in e ences abou hei e ec s on ea nings inequali y. Finally, we use he RIF decomposi ions o di ide he o e all change in ea nings inequali y in o a composi ion e ec ( he componen due o changes in he dis ibu ion o wo ke cha ac e is ics) and a s uc u e e ec ( he componen due o changes in e u ns o hese cha ac e is ics). We ind ha du ing he pe iod om 2004 o 2012, eal ea nings among paid wo ke s inc eased a all pe cen iles and he pe cen age inc ease was g ea e a lowe pe cen iles. Consequen ly, ea nings inequali y declined in u al India. The RIF decomposi ions e eal ha h oughou he ea nings dis ibu ion, excep a he e y op, bo h he compos- i ion e ec and he s uc u e e ec inc eased ea nings, wi h changes in he la e ha ing played a bigge ole. Decomposi ions o inequali y measu es e eal ha in spi e o he composi ion e ec ha ing had an inequali y-inc easing ole, inequali y ell because wo ke s a lowe quan iles expe ienced g ea e imp o emen s in e u ns o hei cha ac e is ics han hose a he op. Ea nings inequali y inc eased as wo ke s acqui ed highe le els o educa ion. A he same ime, lowe e u ns o highe educa- ion educed inequali y. The es o he pape is o ganized as ollows. Sec ion 2 discusses he me hodology used o analyze he change in ea nings. Sec ion 3 desc ibes he da a and he analysis sample. Sec ion 4 p esen s he esul s, and Sec ion 5 concludes. Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 2 o 26 2 Me hodology We b ie ly explain he RIF eg ession o uncondi ional quan iles, ollowed by he RIF decomposi ion echnique. Fo a de ailed exposi ion o his and o he decomposi ion echniques, see Fo in e al. 2011. 2.1 Uncondi ional quan ile eg essions Uncondi ional quan ile eg essions (UQR) in oduced by Fi po e al. (2009) help us exam- ine he ma ginal e ec s o co a ia es on he uncondi ional quan iles o an ou come a i- able. UQR di e om he adi ional quan ile eg essions (Koenke and Basse 1978) in ha he la e examine he ma ginal e ec s on he condi ional quan iles. Fo ins ance, i we obse e ha he condi ional quan ile eg ession coe icien s o college educa ion in- c ease as we mo e om he i s o he nin h decile, we can say ha ha ing mo e people wi h a college educa ion would inc ease ea nings dispe sion wi hin a g oup o indi iduals ha ing he same ec o o co a ia e alues. Howe e , in o de o claim ha college educa- ion inc eases o e all ea nings dispe sion (among all indi iduals i espec i e o hei co a ia es), we need o ely on uncondi ional quan ile eg essions. To unde s and UQRs, we begin wi h he concep o an in luence unc ion (IF). The IF o any dis ibu ional s a is ic ep esen s he in luence o an obse a ion on ha s a is ic. Speci ically, le wdeno e ea nings and le q θ deno e he θ h quan ile o he uncondi ional ea nings dis ibu ion. Then, IF w;qθ ðÞ¼θ−Iw≤qθ g ðÞ= wqθ ðÞ ð1Þ whe e I: g is an indica o unc ion and w is he densi y o he ma ginal dis ibu ion o ea nings. The RIF is ob ained by adding back he s a is ic o he IF. Thus, he RIF o he θ h quan ile is gi en by: RIF w;qθ ðÞ¼qθþIF w;qθ ðÞ¼qθþθ−Iw≤qθ g ðÞ= wqθ ðÞ ð2Þ No e ha he expec ed alue o he RIF is q θ i sel . The condi ional expec a ion o he RIF modelled as a unc ion o ce ain explana o y a iables, X, gi es us he UQR o RIF eg ession model: ERIF w;qθ ðÞjX½¼mθXðÞ ð3Þ In i s simples o m, ERIF w;qθ ðÞjX½¼Xβð4Þ whe e β ep esen s he ma ginal e ec o Xon he θ h quan ile. βcan be es ima ed by o dina y leas squa es (OLS) whe ein he dependen a iable is eplaced by he es i- ma ed RIF. The RIF is es ima ed by plugging he sample quan ile, b qθ, and he empi ical densi y, d wqθ ðÞ, he la e es ima ed using ke nel me hods, in Eq. (2). 2.2 RIF decomposi ion The RIF decomposi ion di ides he o e all change in any dis ibu ional s a is ic in o a s uc u e e ec (due o he changes in e u ns o cha ac e is ics/co a ia es) and a com- posi ion e ec (due o he changes in he dis ibu ion o co a ia es). Compa ed o o he decomposi ion me hods such as he Machado-Ma a (Machado and Ma a 2005), he RIF decomposi ion has he added ad an age o u he di iding he s uc u e and composi ion Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 3 o 26 e ec s in o he con ibu ion o each co a ia e. In his way, i is closes in spi i o he decomposi ion me hod p oposed by Blinde (1973) and Oaxaca (1973). In he case o quan iles, he RIF decomposi ion is ca ied ou using he es ima ed UQR/RIF eg ession coe icien s explained in Sec ion 2.1. The RIF eg ession coe i- cien s o each yea (T) a e gi en by: ^ βT;θ¼Xi∈TXTi⋅X′ Ti  −1Xi∈Td RIF wTi;qTθ ðÞ⋅Xi;T¼1;2ð5Þ The agg ega e decomposi ion o any uncondi ional quan ile θis gi en by: ^ Δθ To al ¼X2 ^ β2;θ− ^ β1;θ  |fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl} ^ Δθ S uc u e þðX2−X1Þ ^ β1;θ |fflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflffl} ^ Δθ Composi ion ð6Þ To examine he con ibu ion o each co a ia e, he wo e ms in (6) can be u he w i en as: ^ Δθ Composi ion ¼Xk¼1 KX2k−X1k  ^ β1k;θð7Þ ^ Δθ S uc u e ¼Xk¼0 KX2k ^ β2k;θ− ^ β1k;θ  ð8Þ Equa ions (7) and (8) ep esen he de ailed decomposi ions o he composi ion and s uc u e e ec s, espec i ely. The de ailed decomposi ion o he s uc u e e ec has a limi a ion when ca ego ical a iables a e included as co a ia es. The choice o he omi ed o e e ence g oup ( o cas e, educa ion, indus y, occupa ion, o s a e o esidence in ou analysis) can in lu- ence he con ibu ion o each co a ia e o he s uc u e e ec . Since he choice o he e e ence ca ego ies is a bi a y, esul s o he de ailed decomposi ion can a y. Exis ing solu ions o he omi ed ca ego y p oblem come a he cos o in e p e abili y (see Fo in e al. 2011). To ensu e he obus ness o ou esul s ega ding he con ibu ion o ac o -speci ic s uc u e e ec s, we use se e al speci ica ions, each o which uses a di e en se o omi ed ca ego ies o he ca ego ical a iables. Though he abo e discussion on RIF decomposi ion ocused on quan iles, i is also applicable o any o he dis ibu ional s a is ic. We p esen he RIF decomposi ion o quan iles as well as selec ed inequali y measu es including he Gini. 3 Da a We use wo ounds o he na ionally ep esen a i e Employmen Unemploymen Su ey (EUS) conduc ed by he Na ional Sample Su ey O ganiza ion (NSSO) o he yea s 2004/05 and 2011/12. Ou a ge popula ion is wage ea ne s be ween he ages o 15 and 64 (wo king age), li ing in u al a eas 4 o 23 majo s a es o India. 5 In bo h yea s, wage ea ne s cons i u ed a ound 25 % o he u al wo king age popula- ion. 6 Nominal ea nings a e con e ed in o eal e ms (2004/05 p ices) using consume p ice indices p o ided by he Labou Bu eau, Go e nmen o India. 7 We also im he eal ea nings dis ibu ion o each yea by d opping 0.1 % o obse a ions om he op and he bo om. 8 Ul ima ely, ou analysis sample consis s o , 44,634 wo ke s in 2004/05 and 36,050 in 2011/12. This co esponds o abou 104 million paid wo ke s in 2004/05 and abou 118 million in 2011/12. Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 4 o 26 4 Resul s In his sec ion we p esen ou indings ela ed o he e olu ion o he ea nings dis ibu- ion in u al India be ween 2004/05 and 2011/12. 4.1 Changes in he dis ibu ion o ea nings om paid wo k Figu e 1 p esen s he ke nel densi y es ima es o he log o eal weekly ea nings o 2004/05 and 2011/12. The ea nings densi y o each yea is skewed o he igh imply- ing ha he median ea ning was less han he mean. O e he 7-yea pe iod, he ea n- ings densi y shi ed o he igh and became mo e peaked (less dispe sed). The mean eal weekly ea nings inc eased om 391 o abou 604 upees, while median inc eased om 263 o 457 upees. Fo 2004/05, he all-India u al po e y line (de ined in e ms o minimum consump ion expendi u e needed o mee a speci ied nu i ional and li - ing s anda d) was 447 upees pe capi a pe mon h (Planning Commission 2014). 9 Thus, he mean (median) eal mon hly ea nings was 3.5 (2.4) imes he po e y line, and in 2011/12 i was 5.4 (4.1) imes his alue. 4.1.1 Changes in ea nings inequali y Figu e 2 plo s he eal weekly ea nings (in upees) a each pe cen ile o 2004/05 and 2011/12. A each pe cen ile, ea nings we e highe in 2011/12 han in 2004/05. The gap be ween he wo cu es e eals ha he inc ease in ea nings was, in absolu e e ms (i.e., measu ed in upees), g ea e o highe pe cen iles. Fo ins ance, eal weekly ea nings inc eased by 99 upees a he i s decile, 194 upees a he median, and 307 upees a he nin h decile. Howe e , as seen in Fig. 3, he pe cen age inc ease in ea nings was g ea e a he lowe end o he dis ibu ion. 10 Fo ins ance, ea nings inc eased by 91 % a he i s de- cile, 74 % a he median, and 44 % a he nin h decile. Thus, ea nings inequali y―de ined in ela i e a he han absolu e e ms―declined o e he 7-yea pe iod. Fig. 1 Ea nings densi ies, 2004/05 and 2011/12 Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 5 o 26 Figu e 4 con i ms he decline in ea nings inequali y: I shows ha he Lo enz cu e o weekly ea nings o 2011/12 lies abo e he one o 2004/05, unambiguously indica - ing ha inequali y declined. Table 1 supplemen s Figs. 2, 3, and 4 and shows how a ious summa y measu es o inequali y changed o e ime. The a io o he ( aw) ea nings a he 25 h o he 10 h pe cen ile was s eady a abou 1.52. A he middle o he dis ibu ion, he e was some dec ease in inequali y as measu ed by he 60 h o he 40 h pe cen ile. In con as , he a io a he 90 h o he 75 h pe cen ile ell e y sha ply om 1.72 o 1.53. Thus, i is clea ha he dec ease in inequali y mainly came om changes a he op and middle o he dis ibu ion han om he bo om. The dec ease in inequali y is also e lec ed in he a iance o log ea nings and in he Gini coe icien s. The Gini o eal weekly ea nings ell om 0.462 o 0.396. 11 This is in sha p con- as o he pic u e in u ban India whe e ea nings inequali y emained i ually unchanged Fig. 2 Real weekly ea nings, by pe cen ile, 2004/05 and 2011/12 Fig. 3 Change in log eal weekly ea nings, by pe cen ile, 2004/05 o 2011/2012 Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 6 o 26 o e he pe iod: The Gini o eal weekly ea nings in u ban India was 0.506 in 2004/5 and 0.499 in 2011/12. Jaya aj and Sub amanian (2015) use consump ion expendi u e da a (also om he NSSO) and ind ha be ween 2004/05 and 2009/10, he Gini declined om 0.305 o 0.299 in u al India. Fo u ban India, i inc eased om 0.376 o 0.393. I is no ewo hy ha while hedi ec iono changein u alinequali y ha hey ind using consump ion expendi u e is he same as wha we ind using ea nings, his is no he case o u ban in- equali y. This makes a s ong case o s udying bo h consump ion and ea nings inequali y. 4.1.2 Wage a es o days wo ked: decomposi ion o he a iance in log ea nings So a ou analysis has been abou weekly ea nings. The EUS also collec s da a on he numbe o hal -days wo ked du ing he week. The ollowing equa ions illus a e he decomposi ion o ea nings inequali y as measu ed by he a iance in log ea nings: Weekly ea nings EðÞ¼A e age daily wage a e WðÞNumbe o days wo ked DðÞ ⇒ln EðÞ¼ln WðÞþln DðÞ ⇒Va ln EðÞ½ |fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl} 1 ¼Va ln WðÞ½ |fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl} 2 þVa ln DðÞ½ |fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl} 3 þ2Co a iance ln WðÞ;ln DðÞ½ |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} 4 The decomposi ion ells us how much o he ea nings inequali y (1) is accoun ed by inequali y o wage a es (2), inequali y o wo kdays (3), and he co-mo emen o wage a es and wo kdays (4). We implemen his decomposi ion o bo h yea s and hen cal- cula e he di e ence be ween co esponding e ms. 12 The esul s a e shown in Table 2. Fig. 4 Lo enz cu es o eal weekly ea nings, 2004/05 and 2011/12 Table 1 Inequali y measu es o eal weekly ea nings om paid wo k 2004/05 2011/12 25-10 1.52 1.51 60-40 1.41 1.32 90-75 1.72 1.53 Va iance o log ea nings 0.61 0.48 Gini 0.462 0.396 Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 7 o 26 In bo h yea s, he co a iance be ween wage a es and days wo ked was posi i e im- plying ha highly paid wo ke s wo ked mo e numbe o days. Also, ea nings inequali y was la gely on accoun o inequali y o wages a es a he han inequali y o days wo ked o because highly paid wo ke s also wo ked o a longe ime: O e 70 % o he ea nings inequali y was due o inequali y o wage a es. 13 The las ow o Table 2 p esen s he decomposi ion o decline in ea nings inequali y as seen in he dec ease in he a iance o log ea nings. Abou 50 % o his decline was due o a decline in inequali y o wage a es. The es was due o a dec ease in inequali y o days wo ked (abou 30 %) and a weake ela ionship be ween highly paid wo ke s wo king mo e numbe o days (abou 20 %). 4.2 Uncondi ional quan ile eg ession esul s Be o e mo ing o he eg ession esul s, we p esen some desc ip i e s a is ics in Table 3 o paid wo ke s in u al India. Mean (log) weekly ea nings inc eased o e he pe iod. The a e age age also inc eased by abou 1.7 yea s, pe haps an indica ion o la e en y in o he labo ma ke as mo e people acqui e highe educa ion. The e was also an in- c ease in he sha e o males, ma ied wo ke s, and Muslims. The p opo ion o hose belonging o ST (Scheduled T ibes) and SC (Scheduled Cas es) declined. 14 Educa ion le els ose signi ican ly: The sha e o illi e a es dec eased by a ound 11 pe cen age poin s, while he sha e o each schooling le el, including college educa ion, inc eased. We classi y indus ies in o se en ca ego ies: ag icul u e, manu ac u ing (including mining), cons uc ion, u ili ies, wholesale and e ail ade, public adminis a ion (includ- ing de ense), and o he se ices (including educa ion, heal h, eal es a e, and inance). O e he pe iod, he majo change in he indus ial dis ibu ion came p ima ily om ag icul u e, which saw a 12 pe cen age poin dec ease, and cons uc ion, which saw a oughly equi alen inc ease. 15 Nex , we es ima e ea nings eg essions (bo h OLS and UQR) sepa a ely o he yea s 2004/05 and 2011/12 wi h he log o eal weekly ea nings as he dependen a iable. The co a ia es include all cha ac e is ics showninTable3and hes a eo esidence. 16 Age en- e s he eg essionsinaquad a ic o masap oxy o wo kexpe ience.“O he s”, and illi e - a es, a e he omi ed ca ego ies o cas e and educa ion, espec i ely. Ag icul u e, and labo e s and unskilled wo ke s, a e he omi ed ca ego ies o indus y and occupa ion, e- spec i ely. Figu es 5 and 6 plo eg ession coe icien s o selec co a ia es. The le column o plo s is o 2004/05 and he igh o 2011/12. Fo each selec ed co a ia e, UQR eg es- sion coe icien s a e plo ed agains he co esponding nine deciles. The dashed lines ep e- sen he 95 % con idence in e al o he coe icien s. The solid ho izon al line is he OLS coe icien . As we mo e ac oss deciles, whe he coe icien s o a pa icula cha ac e is ic a e inc easing o dec easing e eals he e ec o changing he cha ac e is ic on wage inequali y. An upwa d slope sugges s ha inc easing he sha e o wo ke s wi h ha Table 2 Decomposi ion o ea nings inequali y Va [ln(E)] Va [ln(W)] Va [ln(D)] 2 ∗Co [ln(W), ln(D)] 2004/05 0.61 0.43 0.13 0.06 2011/12 0.48 0.36 0.09 0.03 Change o e ime −0.14 −0.07 −0.04 −0.03 Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 8 o 26 Changes in he composi ion o and e u ns o indus y and occupa ion may be pa ly d i en by changes in educa ion. To ha ex en , we should no be including hem as con ols i we a e in e es ed in s udying he o e all con ibu ion o educa ion. Following he decompos- i ion li e a u e, we also es ima e Table 4 wi hou indus y and occupa ion con ols. The e- sul s a e in Appendix 1. 23 Compa ing wi h Table 4, one majo di e ence wi h ega d o he composi ion e ec is ha wi hou indus y and occupa ion con ols, he change in dis ibu- ion o educa ion plays a signi ican ole e en in he bo om o he dis ibu ion (as seen by he 50-10 measu e). O he wise, he conclusions a e quali a i ely he same. 4.3.3 De ailed decomposi ion o he s uc u e e ec The bo om panel o Table 4 p esen s he decomposi ion o he s uc u e e ec . Bo h he 90-10 and he Gini decomposi ions e eal ha educa ion, occupa ion, and being ma ied we e la gely esponsible o he nega i e s uc u e e ec . Fu he , compa ing he 50-10 and 90-50 measu es shows ha o all h ee cha ac e is ics, i was changes in e u ns a he op end o he dis ibu ion ha mainly con ibu ed o he o e all nega i e s uc u e e ec . This was also no ed in Fig. 6 whe e he e u ns o educa ion (wi h illi e a es as he base ca ego y) ac ually declined a he highe end o he wage dis ibu ion, whe eas e u ns did no change signi ican ly in he middle. The same is ue o he e u n o highe occupa ions (wi h labo e s and unskilled wo ke s as he base ca ego y). Compa ing wi h Appendix 1 (wi hou indus y and occupa ion con ols), he conclusions b oadly emain he same. Table 4 Decomposi ion o changes in inequali y measu es om 2004/05 o 2011/12 (Con inued) Ma ied −0.103** 0.013 −0.116*** −0.009 (0.043) (0.032) (0.033) (0.006) Cas e 0.064 0.012 0.052 0.005 (0.055) (0.031) (0.047) (0.011) Muslim 0.013* 0.004 0.010* 0.000 (0.007) (0.004) (0.006) (0.001) Cons an a 0.177 0.114 0.062 −0.094*** (0.256) (0.208) (0.178) (0.035) Boo s ap s anda d e o s shown in pa en heses. Signi ican a *10 %, **5 %, and ***1 % a The cons an does no con ain esidual noise (de ined as di e ence be ween ac ual and p edic ed s a is ics). This is because he RIF me hod decomposes he change o e ime be ween p edic ed s a is ics (see Eq. (6)) Fig. 8 De ailed decomposi ion o he composi ion e ec o selec co a ia es Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 15 o 26 The con ibu ion o e u ns o indus y in Table 4 is in e es ing: i changed in such a manne ha i had an inequali y dec easing e ec a he bo om and an inequali y- inc easing e ec a he op as seen by he nega i e and posi i e e ec s o he 50-10 and 90-50 measu es, espec i ely. I is he e o e no su p ising ha i has an insigni i- can con ibu ion owa d he 90-10 measu e. In Table 4, he con ibu ion o he “cons an ” e m o he o e all s uc u e e ec is la ge and s a is ically signi ican . I is ha d o gi e a meaning ul in e p e a ion o i as i depends on he choice o omi ed ca ego ies o ca ego ical a iables. As desc ibed in Sec ion 2.2, he choice o omi ed ca ego y a ec s he decomposi ion o he s uc u e e ec . We es o he sensi i i y o ou esul s is-à- is choice o omi ed ca ego ies by e-es ima ing Table 4 using wo addi ional speci ica ions p esen ed in Appendix 2. Gi en ha e u ns o educa ion we e la gely d i ing he s uc u e e ec , in he i s speci ica ion we change he omi ed ca ego y o educa ion om illi e a es o he highes educa ional ca ego y, namely, “college and beyond”. As seen in Appendix 2, he e u ns o educa ion a e now posi i e ( is-à- is college and beyond) and he cons an e m is now nega i e. The b oad conclusions a e he e o e he same. In he second speci ica ion, we con e all ca ego ical a iables in o dummy a iables by de ining he a iable o be “0” o he omi ed ca ego y and de ining i o be “1” o he emaining ca ego ies. 24 Educa ion con inues o explain a la ge pa o he composi ion and s uc u e e ec s. 4.4 Robus ness check using s a e po e y lines Recall ha we used he Consume P ice Index –Ru al Labou e s (CPI-RL) o de- la e nominal ea nings o 2004/05 p ices. These p ice indices do no accoun o spa ial p ice adjus men ac oss s a es. As a obus ness check, we use s a e-le el po e y lines compu ed using he Tendulka me hodology (Planning Commission 2014) which accoun o spa ial a ia ion ac oss s a es. We eplica e Tables 1 and 4 using s a e-le el po e y lines and p esen hem in Appendix 3. Ou esul s a e o- bus o he choice o de la o s. 5 Conclusions Using na ionally ep esen a i e da a om he Employmen Unemploymen Su ey, we examine he changes in eal weekly ea nings om paid wo k o u al India om 2004/05 o 2011/12. Fo wage ea ne s who cons i u ed abou a qua e o he u al wo king age popula ion, we ind ha hei eal ea nings inc eased a all pe cen iles. Using consump ion expendi u e da a ha span he en i e popula ion, o he s udies 25 ha e also documen ed an imp o emen in all pa s o he dis ibu ion. Taken oge he , he e is clea e idence ha economic g ow h in he pos - e o m pe iod (a e he ea ly 1990s) has been accompanied by a educ ion in po e y. 26 A he same ime, acco ding o o icial es ima es, in 2011/12, 25.7 % o he u al popula ion was below he po e y line. This igu e ep esen s abou 216.7 million poo pe - sons, a la ge numbe o people li ing below a minimum accep able s anda d. 27 Ou analysis also e eals ha ea nings inequali y in u al India dec eased o e he 7- yea pe iod, and abou hal o he decline can be accoun ed o by he decline in daily wage inequali y. Howe e , while he u al Gini ell o e his pe iod, i emained i ually unchanged in u ban India. This sugges s ha he dynamics o ea nings is di e en o Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 16 o 26 he wo sec o s. This could be because he unde lying s uc u al cha ac e is ics a e di - e en ac oss he wo sec o s. Fo example, while ag icul u e is he la ges employe in u al India, o u ban India i is se ices. I could also be he esul o di e en edis- ibu i e policies ollowed in he wo sec o s. These aspec s need o be ecognized when designing u u e policies o ackle inequali y in he wo egions. Agg ega e decomposi ions o he change in inequali y measu es e eal ha he change in e u ns o wo ke cha ac e is ics was mainly esponsible o he dec ease in ea nings inequali y. Fu he de ailed decomposi ions e eal ha highe le els o educa ion in he popula ion con ibu ed o an inc ease in ea nings inequali y, while lowe e u ns o highe educa ion con ibu ed o a dec ease. Ru al India expe ienced a cons uc ion boom du ing his pe iod ha also con ibu ed o he dec ease in ea nings inequali y. Some s udies (Da e al. 2016; Thomas 2015) ha e a ibu ed he igh ening o he u al casual labo ma ke be ween 2000 and 2012 o he expansion o schooling and o he con- s uc ion boom. O he s (Azam 2012b; Be g e al. 2015; Imbe and Papp 2015) ha e ound ha he MGNREGS (Maha ma Gandhi Na ional Ru al Employmen Gua an ee Scheme), a la ge-scale employmen gua an ee scheme ini ia ed in u al India in 2005, led o an inc ease in casual wages. One canno be ce ain ha his end o ising casual wages and declining ea nings inequali y will con inue in o he u u e. Rega dless o he unde lying causes o he e- cen decline in ea nings inequali y in u al India, ola ili y in global c op p ices and he d ough condi ions cu en ly expe ienced by la ge pa s o he coun y because o wo consecu i e weak monsoons a e impo an eminde s ha policies designed o os e employmen oppo uni ies and wage g ow h o unskilled wo ke s ou side o ag icul u e a e c ucial o imp o ing he economic wellbeing o he second pa o India. Finally, we end wi h he ca ea ha al hough India has he lowes Gini alue among he BRICS coun ies, 28 and we ind ha ea nings inequali y declined in u al India be ween 2004/05 and 2011/12, hese ac s mask ex eme dep i a ions and inequi ies in access o heal h ca e, educa ion, and physical in as uc u e such as sa e wa e and sani a ion (D èze and Sen 2013). One needs o be cognizan ha ex eme inequali ies p e ail in many o he dimensions beyond ea nings and con- sump ion expendi u e. Endno es 1 A no able excep ion is Du a (2005). Fo he pe iod, 1983–1999, a he all-India le el, she inds an inc ease in wage a e inequali y among egula sala ied wo ke s, bu adec ease among casual labo . 2 The e a e some ad an ages in looking a consump ion expendi u e ins ead o ea nings (Goldbe g and Pa cnik 2007). The o me a e a be e measu e o li e ime wellbeing and su e om ewe epo ing e o s. In spi e o his, we eel ha i is impo an o jux apose he wo o ge a comple e pic u e. This is especially impo an as he wo measu es may exhibi di e en ends. K uege and Pe i (2006) documen his o he USA and hen de- elop a model o show how income inequali y can a ec consump ion inequali y. 3 I is ha d o es ablish he supe io i y o one app oach o e he o he . In he SDI amewo k, changes in supply (changes in employmen sha es o demog aphics g oups) and demand (changes in indus ial composi ion) a e assumed exogenous and he e o e una ec ed by changes in he ela i e wage s uc u e. In he RIF decomposi ion, he Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 17 o 26 eedback be ween changing cha ac e is ics and changing e u ns is igno ed. Bo h hese assump ions igno e gene al equilib ium e ec s. 4 In 2004/05, 75.3 % o India’s wo king age popula ion li ed in u al a eas, while in 2011/12 his igu e was 71.1 %. 5 In 2004/05 India had 28 s a es and 7 union e i o ies. We excluded he s a es and union e i o ies o which he e we e no p ice de la o s. The 23 included s a es a e Andh a P adesh, Assam, Biha , Chha isga h, Guja a , Ha yana, Himachal P adesh, Jammu and Kashmi , Jha khand, Ka na aka, Ke ala, Madhya P adesh, Maha ash a, Manipu , Meghalaya, O issa, Punjab, Rajas han, Tamil Nadu, T ipu a, U a P adesh, U a anchal, and Wes Bengal. In bo h yea s, hey cons i u ed 99.3 % o India’s u al wo king age popula ion. 6 In 2011/12, o he emaining u al wo king age popula ion, 30 % we e sel -employed, 2 % we e unemployed, and 43 % we e no in he labo o ce. The main eason o es ic ing ou analysis o wage ea ne s is ha he EUS does no collec ea nings da a o sel -employed indi iduals. Kijima (2006) impu es he ea nings o he sel -employed using Mince ian equa ions es ima ed on he sample o egula wage/sala ied wo ke s. We e ain om his impu a ion as i imposes iden ical e u ns o co a ia es o bo h se s o wo ke s, an assump ion ha may no be ue. 7 We use he Consume P ice Index –Ru al Labou e s (CPI-RL), he ele an p ice index o u al a eas. 8 While we a e awa e ha his may unde es ima e ou inequali y measu es, we do his in o de o emo e po en ial da a en y e o s. 9 The po e y line is based on he me hodology p oposed by he Tendulka Commi - ee in 2009. The commi ee was appoin ed by he Planning Commission, Go e nmen o India. 10 Using consump ion expendi u e da a (also collec ed by he NSSO), o he pe iod be ween 2004/05 and 2009/10, Jaya aj and Sub amanian (2015) ind a simila pa e n o an inc ease in eal consump ion expendi u es a all deciles o u al India, wi h he highes g ow h occu ing a he hi d and ou h deciles. 11 I we conside daily wage a es ins ead o eal weekly ea nings, he Gini ell om 0.398 o 0.358. This indica es ha i is wage a es, and no so much he ime spen wo king, ha is d i ing he dec ease in ea nings inequali y. We s udy his in de ail in he nex sub-sec ion whe e we show he same esul by decomposing he a iance in log ea nings. 12 Al hough he a iance o log weekly ea nings allows us o quan i y a “wage a e e - ec ”,a“wo kday e ec ”, and a “co a iance e ec ”, i does no necessa ily all when one upee is ans e ed om a ich wo ke o a poo one. Howe e , his limi a ion is incon- sequen ial since we ha e shown (using he Lo enz cu es) ha inequali y has unam- biguously allen o e ime. 13 Admi edly, as he e a e bounds o he numbe o days wo ked, anging om hal a day o 7 days, his may ha e pa ly con ibu ed o he lowe inequali y o days wo ked. 14 Scheduled Cas es and T ibes (SC and ST, espec i ely) a e adminis a i e ca ego ies and ep esen g oups o cas es and ibes ha a e en i led o bene i s om a i ma i e ac ion policies such as ese a ions in educa ional ins i u ions and go e nmen jobs o o e come his o ical social and economic disc imina ion agains hem. OBC s ands o O he Backwa d Classes and is a collec i e e m used by he Go e nmen o India o Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 18 o 26 classi y o he cas es ha a e socially and educa ionally backwa d ( o de ails on he cas e sys em, see Deshpande 2011). 15 This shi in indus ial dis ibu ion in u al India has been documen ed in se e al o he s udies including Thomas 2015 and Jacoby and Dasgup a 2015. 16 Following he li e a u e on ea nings eg essions, we also es ima ed he eg essions and decomposi ions wi hou he indus y and occupa ion con ols. The esul s a e quali a i ely he same and a e a ailable om he au ho s on eques . 17 Deshpande e al. (2015) also ind a s icky loo o 1999/2000 and 2009/10 among egula sala ied wo ke s in India. 18 The “O he s”g oup includes, bu is no con ined o, he Hindu uppe cas es as he EUS da a do no allow us o isola e he Hindu uppe cas es. Consequen ly, his ou - way di ision unde s a es he gaps be ween he Hindu uppe cas es and he mos ma - ginalized ST and SC g oups (Deshpande 2011). 19 This inding o u al India is simila o he e idence p esen ed in Azam 2012a o egula sala ied wo ke s in u ban India. Using condi ional quan ile eg essions on EUS da a o 1983, 1993/94, and 2004/05, he inds ha e u ns o seconda y and e ia y educa ion ha e inc eased o e ime and a e la ge a highe quan iles. 20 The esul s based on he o he coun e ac ual ha elies on he cha ac e is ics o 2011/12 and e u ns o 2004/05 a e e y simila and a e a ailable on eques . 21 We also implemen ed he agg ega e decomposi ion using Melly’s e inemen (Melly 2006) o he Machado-Ma a Decomposi ion (Machado and Ma a 2005) and ound simi- la esul s. 22 S anda d e o s o Table 4 (and o all i s a ian s in a ious appendices) we e cal- cula ed using 1000 eplica ions o he boo s ap p ocedu e ollowed by Fo in e al (2011). The basic codes o his a e a ailable om Fo in’s websi e h p:// acul y.a - s.ubc.ca/n o in/da ahead.h ml and we e sui ably modi ied o his pape . 23 We decided o p esen he decomposi ion wi h indus y and occupa ion con ols in he main ex because as no ed ea lie he e was a massi e shi om ag icul u e o in- dus y which we belie e was la gely exogenous o educa ion. Because his change has been widely discussed in ela ed li e a u e on he Indian economy, we eel ha eade s may be mo e in e es ed in he speci ica ion ha includes indus y and occupa ion con- ols, despi e he endogenei y issue ha i su e s om. 24 We had o exclude con ols o s a e o esidence, as he e is no na u al c i e ia o classi ying he s a es as high o low. 25 Ko wal e al. 2011, o all-India, 1983–2004/05; Jaya aj and Sub amanian 2015, o u al and u ban sepa a ely, 2004/05–2009/10. 26 Using NSS da a on consump ion expendi u e om 1957 o 2012, Da e al. (2016) p o ide di ec e idence ha g ow h in India has been accompanied wi h a decline in po e y, especially a e economic e o ms we e ini ia ed in he ea ly 1990s. 27 The co esponding igu es o below po e y line popula ion in u ban India a e 13.7 % (53.1 million). 28 Acco ding o es ima es om he Wo ld Bank, he Gini alues o BRICS coun- ies a e as ollows: B azil-0.539 (2009); Russia-0.397 (2009); India-0.339 (2009); China-0.421 (2010), and Sou h A ica-0.630 (2008). These a e a ailable a Gini Index (Wo ld Bank Es ima e) h p://da a.wo ldbank.o g/indica o /SI.POV.GINI. Accessed on June 1, 2016. Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 19 o 26 Appendix 1 5.1 Re-es ima ing Table 4 wi hou con ols o indus y and occupa ion Table 5 Decomposi ion o changes in inequali y measu es om 2004/05 o 2011/12 No indus y and occupa ion con ols 90-10 50-10 90-50 Gini Value in 2004/05 1.865*** 0.887*** 0.977*** 0.462*** (0.019) (0.017) (0.011) (0.002) Value in 2011/12 1.575*** 0.791*** 0.784*** 0.396*** (0.017) (0.011) (0.015) (0.003) To al change −0.290*** −0.096*** −0.194*** −0.066*** (0.025) (0.020) (0.019) (0.004) Agg ega e decomposi ion o o al change S uc u e e ec −0.406*** −0.105*** −0.301*** −0.095*** (0.027) (0.022) (0.019) (0.004) Composi ion e ec 0.116*** 0.008 0.107*** 0.029*** (0.016) (0.006) (0.015) (0.002) De ailed decomposi ion o he composi ion e ec Educa ion 0.100*** 0.010*** 0.090*** 0.024*** (0.011) (0.003) (0.011) (0.002) Expe ience 0.027*** 0.001 0.027*** 0.008*** (0.004) (0.003) (0.004) (0.001) Male −0.017*** −0.013*** −0.005*** −0.005*** (0.003) (0.003) (0.002) (0.001) S a es 0.003 0.009*** −0.006 0.002*** (0.004) (0.002) (0.004) (0.001) Ma ied 0.003** 0.001 0.001* 0.001*** (0.001) (0.001) (0.001) (0.000) Cas e −0.001 −0.001 −0.001 −0.001** (0.002) (0.001) (0.001) (0.000) Muslim 0.002 0.001 0.001 0.000 (0.001) (0.001) (0.001) (0.000) De ailed decomposi ion o he s uc u e e ec Educa ion −0.244*** −0.042 −0.203*** −0.030*** (0.034) (0.018) (0.030) (0.004) Expe ience −0.360 −0.172 −0.188 0.033 (0.222) (0.174) (0.161) (0.034) Male 0.055 −0.066 0.120*** −0.001 (0.052) (0.048) (0.027) (0.006) S a es −0.085 0.074 −0.159*** −0.007 (0.074) (0.053) (0.053) (0.014) Ma ied −0.105** 0.010 −0.116*** −0.011* (0.044) (0.032) (0.034) (0.006) Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 20 o 26 Appendix 2 5.2 Sensi i i y checks o choice o omi ed ca ego ies Table 5 Decomposi ion o changes in inequali y measu es om 2004/05 o 2011/12 (Con inued) Cas e 0.096 0.018 0.079 0.007 (0.061) (0.032) (0.051) (0.011) Muslim 0.015* 0.004 0.011* 0.000 (0.008) (0.004) (0.007) (0.001) Cons an 0.225 0.069 0.155 −0.087** (0.257) (0.202) (0.188) (0.037) Boo s ap s anda d e o s shown in pa en heses. Signi ican a *10 %, **5 %, and ***1 %. No e ha he i s panel o his able does no exac ly ma ch ha o Table 4. This is because he o al numbe o obse a ions di e ac oss he wo ables as indus y and occupa ion in o ma ion was missing o some indi iduals Table 6 Decomposi ion o changes in inequali y measu es om 2004/05 o 2011/12 Wi h college and beyond as omi ed (all else he same) 90-10 50-10 90-50 Gini Value in 2004/05 1.857*** 0.880*** 0.977*** 0.461*** (0.021) (0.019) (0.011) (0.002) Value in 2011/12 1.576*** 0.791*** 0.784*** 0.396*** (0.017) (0.011) (0.015) (0.003) To al change −0.282*** −0.089*** −0.192*** −0.066*** (0.026) (0.021) (0.018) (0.004) Agg ega e decomposi ion o o al change S uc u e e ec −0.322*** −0.094*** −0.228*** −0.072*** (0.027) (0.024) (0.018) (0.003) Composi ion e ec 0.041** 0.005 0.036*** 0.007*** (0.016) (0.009) (0.014) (0.003) De ailed decomposi ion o he composi ion e ec Educa ion 0.041*** 0.002 0.039*** 0.011*** (0.008) (0.004) (0.007) (0.001) Indus y −0.018** 0.002 −0.020*** −0.009*** (0.008) (0.005) (0.008) (0.001) Expe ience 0.017*** 0.000 0.017*** 0.005*** (0.004) (0.003) (0.003) (0.001) Male −0.014*** −0.012*** −0.002 −0.004*** (0.003) (0.003) (0.002) (0.001) Occupa ion 0.012*** 0.002** 0.010*** 0.003*** (0.004) (0.001) (0.004) (0.001) S a es 0.001 0.009*** −0.009 0.002** (0.015) (0.002) (0.014) (0.001) Ma ied 0.002* 0.001 0.001 0.001*** (0.001) (0.001) (0.001) (0.000) Cas e −0.002 −0.001 −0.001 −0.001*** (0.001) (0.001) (0.001) (0.000) Muslim 0.002 0.001 0.001 0.000 (0.001) (0.001) (0.001) (0.000) Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 21 o 26 Table 6 Decomposi ion o changes in inequali y measu es om 2004/05 o 2011/12 (Con inued) De ailed decomposi ion o he s uc u e e ec Educa ion 0.497*** 0.086 0.411** 0.098*** (0.192) (0.055) (0.185) (0.035) Indus y −0.006 −0.061*** 0.055** 0.009* (0.028) (0.017) (0.024) (0.005) Expe ience −0.313 −0.185 −0.128 0.025 (0.223) (0.173) (0.156) (0.033) Male 0.084 −0.051 0.135*** 0.005 (0.053) (0.050) (0.028) (0.006) Occupa ion −0.088*** 0.008 −0.096*** −0.012*** (0.028) (0.023) (0.026) (0.004) S a es −0.01 0.085 −0.094** 0.014 (0.069) (0.052) (0.048) (0.013) Ma ied −0.103** 0.013 −0.116*** −0.009 (0.043) (0.032) (0.032) (0.006) Cas e 0.064 0.012 0.052 0.005 (0.056) (0.030) (0.047) (0.011) Muslim 0.013 0.004 0.010 0.000 (0.01) (0.005) (0.011) (0.001) Cons an −0.461 −0.004 −0.456* −0.208*** (0.307) (0.206) (0.235) (0.052) Boo s ap s anda d e o s shown in pa en heses. Signi ican a *10 %, **5 %, and ***1 % Table 7 Decomposi ion o changes in inequali y measu es om 2004/05 o 2011/12 All ca ego ical a iables con e ed o dummy a iables, excluded s a es as con ols 90-10 50-10 90-50 Gini Value in 2004/05 1.857*** 0.880*** 0.977*** 0.461*** (0.021) (0.019) (0.011) (0.002) Value in 2011/12 1.576*** 0.791*** 0.784*** 0.396*** (0.017) (0.010) (0.015) (0.003) To al change −0.282*** −0.089*** −0.192*** −0.066*** (0.027) (0.022) (0.019) (0.004) Agg ega e decomposi ion o o al change S uc u e e ec −0.382*** −0.092*** −0.290*** −0.082*** (0.028) (0.024) (0.019) (0.005) Composi ion e ec 0.101*** 0.003 0.098*** 0.017*** (0.015) (0.007) (0.013) (0.002) De ailed decomposi ion o he composi ion e ec Educa ion 0.046*** 0.000 0.047*** 0.010*** (0.006) (0.003) (0.005) (0.001) Indus y 0.026*** 0.009** 0.017*** 0.001 (0.006) (0.004) (0.005) (0.001) Expe ience 0.039*** 0.001 0.038*** 0.009*** (0.005) (0.003) (0.004) (0.001) Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 22 o 26 Appendix 3 5.3 Robus ness check using s a e-le el po e y lines as de la o s Table 7 Decomposi ion o changes in inequali y measu es om 2004/05 o 2011/12 (Con inued) Male −0.023*** −0.011*** −0.011*** −0.006*** (0.004) (0.003) (0.002) (0.001) Occupa ion 0.015*** 0.002** 0.012*** 0.003*** (0.005) (0.001) (0.004) (0.001) Ma ied 0.001 0.001 0.000 0.001*** (0.001) (0.001) (0.001) (0.000) Cas e −0.002* 0.000 −0.002* −0.001* (0.001) (0.000) (0.001) (0.000) Muslim −0.001 0.001 −0.002* −0.001*** (0.001) (0.001) (0.001) (0.000) De ailed decomposi ion o he s uc u e e ec Educa ion −0.133*** −0.040** −0.093*** −0.016*** (0.030) (0.017) (0.026) (0.003) Indus y −0.060** −0.057*** −0.003 −0.002 (0.024) (0.016) (0.021) (0.003) Expe ience −0.391 −0.189 −0.202 0.022 (0.242) (0.173) (0.190) (0.036) Male 0.066 −0.044 0.110*** 0.007 (0.054) (0.051) (0.026) (0.005) Occupa ion −0.042 0.005 −0.047* 0.001 (0.028) (0.011) (0.026) (0.004) Ma ied −0.160*** 0.005 −0.165*** −0.023*** (0.045) (0.032) (0.035) (0.006) Cas e 0.127** 0.013 0.114** 0.006 (0.063) (0.028) (0.056) (0.012) Muslim 0.016** 0.003 0.012** 0.002* (0.007) (0.004) (0.006) (0.001) Cons an 0.196 0.211 −0.015 −0.079** (0.287) (0.203) (0.223) (0.038) Boo s ap s anda d e o s shown in pa en heses. Signi ican a *10 %, **5 %, and ***1 % Table 8 Inequali y measu es o eal weekly ea nings om paid wo k Using s a e po e y lines 2004/05 2011/12 25-10 1.52 1.52 60-40 1.39 1.27 90-75 1.70 1.50 Va iance o log ea nings 0.59 0.46 Gini 0.452 0.387 Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 23 o 26 Table 9 Decomposi ion o changes in inequali y measu es om 2004/05 o 2011/12 Using s a e po e y lines as de la o s 90-10 50-10 90-50 Gini Value in 2004/05 1.797*** 0.864*** 0.933*** 0.451*** (0.015) (0.010) (0.012) (0.002) Value in 2011/12 1.515*** 0.780*** 0.735*** 0.387*** (0.015) (0.012) (0.011) (0.003) To al change −0.282*** −0.084*** −0.198*** −0.065*** (0.021) (0.015) (0.016) (0.004) Agg ega e decomposi ion o o al change S uc u e e ec −0.306*** −0.082*** −0.223*** −0.073*** (0.021) (0.018) (0.015) (0.004) Composi ion e ec 0.024* −0.001 0.025** 0.009*** (0.014) (0.009) (0.012) (0.003) De ailed decomposi ion o he composi ion e ec Educa ion 0.036*** 0.003 0.033*** 0.012*** (0.007) (0.004) (0.006) (0.001) Indus y −0.018** 0.002 −0.020*** −0.008*** (0.007) (0.004) (0.007) (0.001) Expe ience 0.014*** −0.001 0.016*** 0.005*** (0.003) (0.003) (0.003) (0.001) Male −0.021*** −0.016*** −0.006*** −0.004*** (0.004) (0.003) (0.001) (0.001) Occupa ion 0.009*** 0.001 0.009*** 0.003*** (0.003) (0.001) (0.003) (0.001) S a es 0.002 0.009*** −0.006 0.002 (0.015) (0.003) (0.014) (0.001) Ma ied 0.002** 0.002** 0.001 0.001*** (0.001) (0.001) (0.001) (0.000) Cas e −0.003* −0.001 −0.001 −0.001*** (0.001) (0.001) (0.001) (0.000) Muslim 0.002 0.001 0.001 0.000 (0.001) (0.001) (0.001) (0.000) De ailed decomposi ion o he s uc u e e ec Educa ion −0.166*** −0.023 −0.143*** −0.016*** (0.033) (0.030) (0.046) (0.004) Indus y −0.045* −0.050*** 0.005 0.008* (0.027) (0.016) (0.023) (0.004) Expe ience −0.411* −0.175 −0.236 0.021 (0.213) (0.169) (0.148) (0.034) Male −0.047 −0.115*** 0.068*** −0.001 (0.044) (0.042) (0.026) (0.006) Occupa ion −0.144*** −0.003 −0.141*** −0.012*** (0.028) (0.023) (0.026) (0.004) S a es −0.164** −0.146** −0.018 0.013 (0.070) (0.059) (0.047) (0.014) Khanna e al. IZA Jou nal o Labo & De elopmen (2016) 5:18 Page 24 o 26