Income mobility in Argentina

Transcription

1 Inome molty n Argentna Lus Beara and Fernando Grosman 2 Astrat The paper analyses the varalty of laour nomes n Argentna from md-eghtes to The magntude of the nome-nstalty phenomenon s estmated and ts determnants are evaluated under dfferent maroeonom ontexts. The doument also analyses how nome flutuatons have affeted the nequalty of nome dstruton. Fnally, the nome onvergene hypothess s explored. Data for Greater Buenos Ares wll e used; ths s Argentna s man metropoltan area and home to nearly one thrd of the populaton. The analyss, overng the perod wth panel data, wll dstngush four perods that are relatvely homogeneous n terms of a set of varales that are mportant for the ams pursued. The onvergene hypothess wll e tested usng dynam pseudo-panel. One of the man results ndates that the growth n oupatonal nstalty regstered sne the md-nnetes led to a hgh varalty n nomes despte the maroeonom stalty enjoyed throughout the nnetes. In addton, a shft n the haratersts of nome molty was verfed etween the extremes of the perod. Moreover, the panorama of growng nequalty n the dstruton of nome s also approprate to desre what happened wth the hanges n the dstruton of more permanent nomes. Fnally, the results otaned show sare long-term nome molty n Argentna ndatng that the nome path does not onverge to the general mean. Key words: Inome molty, nome nstalty, nome dstruton JEL Classfaton: D3, J6, I3 Unversdad Naonal de General Sarmento, Argentna. lear@ungs.edu.ar 2 Natonal Centre for Sentf and Tehnal Researh (CONICET) and Unversdad Naonal de General Sarmento, Argentna. fgrosma@ungs.edu.ar

2 I. Introduton The dstruton of oth ndvdual and famly nomes n Argentna has eome steadly more onentrated sne the md-970s. Ths trend lasted throughout the 980s, whh were largely years of nstalty and stagnaton, and nto the followng deade, despte etter maroeonom performane. 3 Throughout ths perod, there were also sharp hanges n nflaton: very hgh rates n the 970s and 980s, nludng hypernflatonary spkes towards the end of the latter deade and n the early 990s, efore sustantal pre stalty was restored n the remander of the last deade of the twenteth entury. As nflaton s a key fator n explanng the stalty of real nomes, the latter ought to have worsened n the 970s, and espeally n the 980s, and then should have mproved n the followng deade. Nonetheless, there s also evdene of hgh levels of jo nstalty, partularly n the 990s, 4 whh also affets nome varalty at oth the ndvdual and household levels. It s therefore worth makng a more n-depth analyss of nome nstalty n Argentna s dfferent maroeonom stuatons, gven the adverse effets of suh nstalty on ndvdual welfare levels. In partular, nstalty nreases rsk and thus dmnshes the utlty of a gven flow of resoures; and t an also undermne onsumpton levels even when predtale. Instalty may go hand-n-hand wth molty, whh generally means hanges n the relatve poston of nomes n the dstruton, or hanges n the dfferentals etween them. The exstene of a molty proess has an mpat on the nome dstruton. In partular, t ould make the degree of onentraton, measured y a gven year s nomes, overstate nequalty n the dstruton of more permanent nomes measured as an average over several years. More mportantly for the purposes of many dagnost studes, hanges n the degree of molty may ause hanges n urrent-nome nequalty to nadequately reflet hanges n the nequalty of average nomes. Ths aspet s dretly related to the presene of onvergent or dvergent movements, n so far the former would usually (ut not always) lead to mprovements n ross-seton nequalty. Gven the mportane of varatons n nomes for analysng ther level and dstruton, ths paper wll examne the hanges that have ourred n the dfferene n the dstruton n Argentna sne the md 980s. Despte the mportane of suh ssues, few studes have addressed them n the past. Moreover, the few analyses that have een undertaken 5 use a shorter tmeframe than onsdered here; and, n partular, they do not nlude perods of hgh nflaton. They also fal to expltly relate the phenomena of nstalty, rsk, molty and onentraton n average nomes. The researh summarzed n ths artle studed the key fators determnng nstalty and ts dfferental ntensty etween household groups. It also assessed the extent to whh hanges n nstalty have affeted hanges n the nome dstruton. The am, therefore, s to explore the hypothess that an nrease n nequalty, when studed usng data from eah 3 See, for example, Altmr and Beara (200). 4 See Hopenhayn (200), Galan and Hopenhayn (2000), and Beara and Maurzo (2004). 5 See Alornoz and Menéndez (2002), Crues and Wodon (2003), Gutérrez (2004), and Felds and Sánhez Puerta (2005). 2

3 perod, or ross-seton data (.e. wth urrent nomes) also reflets hanges n the dstruton of average nomes. To omplement ths, average household nome was adjusted for the effet of varalty, and ts ehavour was ompared wth that of unadjusted average nome. The analyss wth longtudnal data, overng the perod , wll dstngush four perods that are relatvely homogeneous n terms of a set of varales that are mportant for the ams eng pursued. Data for Greater Buenos Ares wll e used, sne ths s Argentna s man metropoltan area and home to nearly one thrd of the populaton. The temporal and geograph seton hosen reflet the avalalty of statstal nformaton, sne mrodata are only ontnuously avalale for that regon and for those years of household surveys. The study omts , sne these are dffult years to analyse wth the methodology used n ths paper. 6 The onvergene hypothess wll e tested usng dynam pseudo-panel from 984 to Seton II, whh follows, ontextualzes the analyss of molty n Argentna, y refly summarzng the ehavour of the maroeonomy and the nome dstruton. Seton III revews a numer of the dfferent approahes to e found n the lterature on nome dynams and hghlghts the varous onerns that motvate analyss of ths top. Seton IV sets out the artle s spef ojetves and desres the analytal methods appled; whle seton V desres the data soure used. The ore of the artle onssts of setons VI and VII, whh desre and analyse the fgures for Argentna n terms of varalty and molty, respetvely. Seton VIII offers onlusons. II. Maroeonom ehavour and nome dstruton sne the md-970s The md-970s marked the start of a 5-year perod of maroeonom nstalty and produtve stagnaton. Gross domest produt (GDP) was roadly unhanged throughout that perod, and nflaton remaned at hgh levels (fgure ). Ths performane was assoated wth an external onstrant arsng from the hgh level of external det, whh n turn was generated y the poles mplemented, partularly etween 978 and 98. The measures adopted susequently throughout the 980s were unale to suessfully address a numer of strutural aspets of the Argentne eonomy, suh as the management of pul aounts and the hgh nflaton regme (although the two fators are not ndependent). The latter s very mportant for understandng oth the domest effets of external orrowng and the dffultes n ahevng sustaned stalzaton. 6 As wll e seen elow, nstalty s evaluated from data showng hanges n ndvdual nomes over 8- month perods. To nlude the ntal months of 2002 would e heterogeneous n terms of nflaton, eause they nlude tmes of relatve stalty, therey makng t mpossle to adequately haraterze ths phase. The phase also overs dfferent moments n the dynam of aggregate output. 3

4 Fgure. Greater Buenos Ares: Inequalty and nflaton 0,55 989: 4900% 990: 300% Gn per apta household nome Gn oeffent 0,5 0,45 0,4 0,35 Annual nflaton rate (CPI) Gn personal nomes Annual nflaton rate (Deemer-Deemer) (%) 0, Soure: Authors estmates on the ass of data provded y the Natonal Insttute of Statsts and Censuses (INDEC). - Ths proess of maroeonom nstalty ulmnated n the hypernflatonary epsodes of 989 and 990. The Government that took offe n 989 was ntally unale to mprove the stuaton, and t was left to the eonom team apponted n late January 99 to mplement a stalzaton programme that managed to halt nflaton and generate atvty growth. The ornerstone of that programme was the Convertlty At, whh fxed the exhange rate, estalshed the onvertlty of all urreny n rulaton and prohted monetary ssuane that was not aked y external assets. Ths measure, together wth a numer of others mplemented n the fsal and other domans, allowed for a rapd reduton n nflaton: the varaton n the onsumer pre ndex was rought down to 3% 7 n May 99, and to levels lose to % y the end of that year. Stalty frstly allowed for an mprovement n the purhasng power of wages and, seondly, an expanson of redt. These developments were assoated wth sgnfant onsumpton growth, partularly n the ase of durale goods and onstruton. Investments made y prvatzed enterprses also ontruted to the 7 The rate flutuated around 7% etween Otoer 990 and January 99, efore rsng to 27% n Feruary as a result of the devaluaton and other measures (suh as rate hkes). 4

5 expanson of domest demand, 8 whle the reduton n nflaton made t possle to mprove levels of tax revenue. The vgorous nflow of foregn aptal etween 99 and 994 attrated y the greater onfdene generated y stalty and the orentaton of eonom poly, ut also due to a larger supply of funds on the nternatonal fnanal market oosted the growth of domest demand. Nonetheless, the Mexan rss n late 994 revealed the fraglty of an eonomy n whh expanson was ased on aptal nflows from aroad, although the Argentne reesson assoated wth ths event was ref, and the eonomy resumed a rapd growth path as soon as ondtons on the nternatonal aptal market mproved. In 998, however, when ths market eame more prolemat agan and Brazl (a major export destnaton) went nto reesson, there was a new downswng n GDP whh, unlke the prevous epsode, lasted an unommonly long tme and trggered aandonment of the fxed exhange rate system shortly after the start of The serous maroeonom nstalty experened sne the md-970s s one of the explanatons for the sgnfant deteroraton n the nome dstruton sne that tme. Intally, the nrease n nequalty proaly stemmed from the dfferental mpat of the rse n nflaton n 975 and 976 on the relatve wages of ndvduals wth dfferent skll levels. Inome nequalty remaned hgh n the 980s, wth ndvdual nomes mantanng ther onentraton whle famly nomes eame more onentrated. Despte an mprovement n the maroeonom settng and the ntroduton of strutural reforms, nequalty ontnued to worsen n the 990s, exept durng the ntal expansonary phase (99-994). Although an n-depth analyss of that trend s eyond the sope of ths artle, t s rual to keep n mnd the effets of the sgnfant deteroraton n the laour market. Unemployment, for example, rose from 6% at the start of that deade to 2% n 994 and 8% n 200; and t remaned at 5% even durng the reovery of Ths phenomenon had ts greatest effet on wages, employment possltes and jo qualty, partularly for the lower sklled. One onsequene of the unsatsfatory employment trend n the 990s was an nrease n jo molty. As mentoned n the ntroduton, a numer of studes report an nrease n the qut rate, partularly n non-wage and prearous wage-earnng jos that were not regstered wth soal seurty. An nrease n the proporton of the latter, whh dsplayed less average stalty, was an addtonal fator leadng to a shorter average duraton of jos. III. The dfferent ams of studes on nome dynams Numerous papers have studed hanges n ndvdual and/or famly nomes over tme, usng panel data. Some of these analyse nome nstalty y evaluatng ts ntensty and varaton through tme, or etween groups of ndvduals, or else y nvestgatng the mpat of nstalty on ndvdual and famly welfare. A learly larger volume of researh, however, fouses on nome molty,.e. hanges n the relatve poston of nomes n the dstruton, or the dfferentals etween them, over tme. Suh studes reflet two types of onern: some 8 The prvatzaton proess was mplemented rapdly, eause n addton to supportng the goal of wthdrawng the State from produtve atvty, aptal nflows were essental to support the external and fsal aount alanes. 5

6 nvestgate the magntude and haratersts of molty and how t has hanged over tme, whle others examne the effets of molty on nequalty. 3.. Inome molty Many studes analyse the paths of personal or household nomes wth a vew to evaluatng hanges n ther relatve poston n the dstruton through tme. Changes n the rankng of nome repent unts n the nome dstruton are generally referred to as nome molty. Inome paths an also e traked for the purpose of analysng the dreton and magntude of the hanges they experene, whether or not aompaned y alteratons n the rankng. Ths s known n the spealzed lterature as asolute molty. The two approahes are omplementary and an our smultaneously and wth dfferent ntenstes: for example, hgh/low molty n the rankng an our n onjunton wth a low/hgh asolute molty. Ths depends partly on the nequalty that exsts n the dstruton of urrent nome, eause, when nequalty s hgh, the asolute hange n nome needed to ause a hange n the rankng wll e greater than n a low-onentraton stuaton. The proporton of nome repents that hange ther poston n the dstruton s normally analysed through matres that show transtons from one quantle of the dstruton to another, etween two perods. Although ths s the most ommon proedure n the spealzed lterature, t has lmtatons: n partular, t fals to apture hanges that take plae wthn the s of the seleted quantles. 9 Some authors have tred to orret these shortomngs, e.g. y makng the ares of nome quantles flexle (Hlls, 998). Other ways to otan quanttatve evdene of molty are through measures of assoaton suh as the smple and rank orrelaton oeffents, of Pearson and Spearman, respetvely (OECD, 996). It should e kept n mnd, however, that the frst of these oeffents s not restrted to hanges of rank. Moreover, oth the matres and the orrelaton oeffents an only e used to evaluate hanges etween two perods. They are not sutale for analysng asolute molty, eause transton matres, for example, do not reord nome hanges that do not ental hanges of rank. Aordngly, when a hange of rank s not a onern, proedures that spefally quantfy the magntude of the hange n nomes are generally used. For example, Felds (2004) uses the followng ndator of asolute varaton, where represents eah nome repent, and y represents nomes n t 0 = and t =2, Ω = n ( / n) y2 y () = although an evaluaton s also made of the dfferenes etween the logarthms of nomes, or a gven nome repent s share of the total. These ndators express the degree of molty 9 It s also a measure that s senstve to the degree of nequalty n the soety, and, therefore, s unsutale for omparson etween ountres. For example, the same 0% nrease n nomes ould represent a quantle hange n one ountry, whereas n another t ould mean stayng n the same nome raket. 6

7 wthout ndatng the dreton of the hange. Inludng the sgn of the hange makes t possle to norporate the dreton of the mean varaton whh s refleted n Ω*. Ω * = n ( / n) ( y2 y ) (2) = A partular onern s to analyse the extent to whh the presene of molty s assoated wth nome onvergene. The latter, also known as mroonvergene, ours when nomes experene hanges that rng them loser to the mean nome n the dstruton. The analyss n ths ase usually fouses on the sgn of the oeffent β n a model suh as: y = + β lny ln α + ε (3) where lny s the dfferene of the logarthm of nome etween two perods, t 0 and t, and lny s the logarthm of the ntal perod. When the oeffent β takes a negatve (postve) value, there wll e onvergene towards (dvergene from) the mean. 0 An advantage of regresson analyss s that t an nlude numerous nome oservatons for eah repent and also evaluate the fators drvng hanges n nomes The mpat of molty on nequalty A seond type of researh, losely related to the am of molty analyss, seeks to evaluate the mpat of hanges n ndvdual nomes on the nome dstruton. In partular, t asks whether the degree of nequalty measured wth ross-seton data dffers, and y how muh, from that orrespondng to permanent nome, measured as average nome over several perods. Muh of the lterature on nome dynams has foused on estmatng the equalzng effet of molty, to otan an approxmate measurement of more permanent nequalty n soety. Suh studes tend to ompare the varous nequalty ndexes alulated from rossseton nome data, wth ndators ased on longer-run nomes. In general, the Shorroks (978) methodology has een used, whh analyses the ntensty of ths equalzng effet through an adjustment of nequalty for molty ndex (R) whh ompares the onentraton of average nome n the perod under onsderaton wth the average of the nequaltes of those varous oservatons. I( w ( ) = R W T T t= ) η I( w ) t t (4) 0 Sometmes quantle regressons are used; see, for example, Fontenay, Gorgens and Lu (2002). 7

8 where I s the nequalty ndator, w s average nome over T perods, w t s nome n perod t, and η t s the weghtng fator defned as the unts share of total nome n perod t wth respet to the nome n the set of T perods. Felds (2004) argues that f the am s to evaluate the extent to whh molty altered the nequalty measured at a gven pont n tme, the omparson should e made dretly etween I (w ) and I ( w T ),.e. etween nequalty n ntal perod and the nequalty of average nome. R tends to zero as a maxmum value when there s no molty, and dereases as the effet of molty on the dstruton ntensfes. One pont to hghlght s that the presene of onvergent nome molty s not neessarly translated nto an mprovement of nome nequalty when the latter s evaluated wth ross-setonal data. Sne ths statement may not seem ntutve, t s worth takng a moment to onsder t further. In the sheme that follows we present four examples that llustrate the possle omnatons of the evoluton of stat nequalty and the onvergene/dvergene. In the lower-left quadrant C t an e oserved that the varaton of repents nomes etween oservatons and 2 was onvergent whle stat nequalty rose. Note that for ths to happen, the fnal nome of at least some of those repents whose nomes have mproved/worsened has to e hgher/lower than the nomes of the repents who were ntally n those postons. Examples of possle omnatons of onvergene/dvergene of nomes and the evoluton of stat nequalty A. Dvergene and nrease of stat nequalty B. Dvergene and reduton of stat nequalty Average of Average of Repents Os Os 2 Os Os 2 os. and 2 os. and 2 A B C Inequa lty (VC) C. Convergene and nrease of stat D. Convergene and reduton of stat nequalty nequalty Average of Average of Repents Os Os 2 Os Os 2 os. and 2 os. and 2 A B C Inequa lty (VC) The tme perod over whh more permanent nomes s alulated matters eause the longer the perod, the smaller one would expet the dfferenes etween average nomes to e. 8

9 3.3. Changes n the ntensty of molty Panel data ontrute to a etter evaluaton of the dynam of nequalty under the hypothess of hanges n the ntensty of molty. But f the latter were onstant, measurements of nequalty usng spef data would adequately reflet what would happen to the dreton of the hange n the onentraton of more permanent nomes. An nrease n stat nequalty wll not generate greater nequty n the dstruton of more permanent nomes only f there s a onomtant nrease n nome molty. Spefally, as shown y Gottshalk and Danzger (998), the varane of average nomes s a funton of the average of the varanes of the dstrutons of eah oservaton and the average of the ovaranes etween the dfferent oservatons The welfare effet of nome nstalty A dfferent onern s to evaluate the ntensty of the nstalty of ndvdual nomes nsofar as ths dmnshes the utlty of a gven volume of eonom resoures. In partular, varalty nreases rsk, 2 and, although t an e antpated, t an also hange utlty, partularly n ountres wth poorly developed redt markets. If two households reeved the same average nome at the end of the year, ut one of them had no nome for half of that year, whereas the other reeved /2 of ts annual nome every month, the welfare levels of the two repents are lkely to have een very dfferent. The evaluaton of nome flutuatons s generally ased on estmatng the degree of ntertemporal varaton around an expeted nome or oserved average nome. The tradtonal ndator for ths purpose s the oeffent of varaton, although the varane or mean devaton of the logarthm of nomes (Gottshalk and Danzger, 998 and Shorroks, 978) are also used. Some authors also use the resduals from fxed-effets wage-regresson models as a measure of varalty (Burgess, Gardner and others, 2000). Other proedures use the varalty of oserved nomes to estmate an nome orreted for flutuatons, whh normally entals applyng rsk funtons to estmate an average nome whh, f fxed, would provde the same utlty as that atually reeved. 3 Ths s ased on the dea that ndvduals (.e. nome repents) are rsk-averse; so, the greater the varalty of nomes the smaller the utlty otaned from them. The utlty funtons used strtly onave are defned y a parameter of averson to varalty, ρ, whh determnes the nstalty dsount suffered y nome repents. υ ( ) An example of a funton that takes aount of rsk averson s the followng: ρ y sρ = ρ lny sρ = y (5) whh shows that the utlty of a gven nome dereases as rsk averson (the oeffent ρ) 2 See Arrow (970). 3 These proedures stem from dstruton analyses ased on the soal welfare approah formulated y Atknson (970). See Cowell (2000). 9

10 rses. In ths ase, orreted nome y (the level of onstant nome that provdes the same utlty as the flow of oserved nomes) s alulated as follows: T ρ [ g() t y ] ρ t y (6) = t= = = T where g () t t Fnally, we examned to what extent nome molty gves rse to onvergent movements. In ths paper we approah ths ssue through the estmaton of regresson ln y = + β lny 0 α + ε where lny s the logarthm of nome of the urrent perod and lny 0 s the logarthm of nome n the prevous perod. There would e onvergene/dvergene n ase β was lower/hgher than. These models were appled to longtudnal data and to data omng from fttous seres. (7) IV. Ojetves and methods Gven the wde range of nterests represented n the lterature on nome dynams that make use of panel data, t s worth larfyng here the spef ams of the researh whose ntal results are presented n ths artle. Frstly, the degree of nstalty of real nomes was analysed, sne ths has an adverse affet on ndvdual and famly welfare. Analyss of nome varalty over short perods s a relatvely unexplored top, proaly eause t s not a sgnfant phenomenon n the world s leadng eonomes. Nonetheless, n ountres suh as Argentna, where maroeonom stalty has een a feature throughout muh of ts modern hstory, nome varalty s partularly relevant, rrespetve of any dstrutve mpats espeally, as wll e seen, when t seems to persst even n stuatons of pre stalty. 0

11 Seondly, the researh evaluated the degree of nome molty, along wth ts mpat on the dstruton of more permanent nomes. Two analytal approahes were used to measure nome nstalty. The frst of these measured the varalty of oserved urrent nomes (of ndvduals and famles) around the mean, usng the oeffent of varaton (CV for ndvduals and CV h for households). CV = T t = ( w w ) w t 2, where w T T w t t= = (7) CV h where = T t = m = ( y y ) y ht h y ht = w ht h 2, T s the numer of oservatons avalale, and dentfes eah of the m household memers who were employed n at least one of the four oservatons. (8) y h T t= = T y ht Mean varalty arses dretly from averagng the CVs of eah ndvdual and household. As the mpat of nstalty was assumed to vary aross nome repents and among famles (greater mpat among less sklled workers and lower-nome famles), dsaggregated estmates were made for oth ases, usng groups ased on the level of shoolng of the ndvdual or head of household as a proxy for sooeonom status. Steps were also taken to otan evdene on the mportane of phenomena dretly related to varalty. For example, nstalty n real ndvdual nomes s assoated wth hanges n hourly pay and wth hanges n oupatonal status (employed/unemployed). The ntensty of the latter wll hange espeally when jo molty vares; whereas varatons n nomnal wages are assoated, among other fators, wth the degree of pre stalty, and are lkely to e larger and more frequent n nflatonary settngs. 4 The tme for whh a person works an also e spefed n greater detal and roken down nto two parts: oupatonal varalty and varalty of the numer of hours a person works whle employed. 4 Nomnal hourly nomes an vary merely as a result of hanges n earnngs from a gven jo; ut they an also vary as a result of movng from one jo to another. The mpat of ths effet was not alulated.

12 To demonstrate the mpat of some of these varales, a proedure was arred out to dentfy the varalty of nomes when the effets of jo nstalty and the varalty of monthly remuneraton (whh therefore also reflet hanges n hours worked) are suessvely solated. In the frst ase, to measure the effet of hanges n remuneraton, the oeffent of varaton of remuneraton from the jos of eah ndvdual s alulated,.e. that arsng from postve nomes only, exludng oservatons orrespondng to stuatons n whh the person was not employed (CV ao ). v ao n t= = * ( w w ) w t * 2 for w t > 0 (9) where n s the numer of oservatons n whh ndvdual has a postve nome (.e. where w t > 0 ) w * n t= = n w t To otan an ndator that solates hanges n remuneraton, a oeffent of varaton s alulated on values whh, when postve, orrespond to the frst oservaton n whh the person was employed. v ar T t = ** ( A w ) w t ** 2 = (0) wt w > 0 wth At = ; 0 wt = 0 w ** T t= = T A t where w represents remuneraton n the frst oservaton wth a postve value. Two fators are assumed to affet the varalty of nomnal household laour nomes: varatons n the numer of nome earners n the household and varalty n the nomes reeved y them. As these two fators an work n opposte dretons they may offset eah other, ether partally or ompletely; n the latter ase the resultant hange n the varalty of household nomes s zero. Varatons n the numer of household nome earners may reflet hanges n the sze of the household, hanges n the employment rate of a household n whh the numer of memers does not hange, or oth fators together. Ths paper does not dstngush etween the auses of suh varaton. 2

13 The magntude of the nstalty of remuneraton s dedued from the oeffent of varaton of the nome of eah household, alulated assumng the numer of employed memers (Cv ao h) s unhanged. In ths ase, household memers who were employed at some pont had an nome mputed to them for perod(s) n whh they were unemployed, equvalent to that reeved n the nearest perod (ether efore or after) n whh ther remuneraton was postve, adjusted for the mean varaton n nomes etween the two perods. CV ao h T * * ( y y ) 2 ht h t= = () * yh where: y T * t= h = T y * ht ; * ht m y = B ; = ht ht w = w ht ht w w ht ht > 0 B ; = 0 n H w t = w ht = h= ; w ht = w hs wt w s wth s eng the postve-nome perod nearest to t, whle w t s the average wage n perod t, n s the numer of persons employed n the perod, and H s the numer of households. To evaluate the sgnfane of hanges n the numer of employed persons, the oeffent of varaton of famly nome was alulated, and t was assumed that the monthly remuneraton of all employed memers remaned fxed and equal to the frst postve oservaton n eah ase (Cv ar h). CV ar h T ** ** ( y y ) 2 ht h t= = ** (2) yh 3

14 where ** n y ht = A jht j= A jht wht wh > 0 = 0 wht = 0 where w h s the remuneraton n the frst perod wth a postve value. The seond analytal approah to nome nstalty reognzes how utlty delnes when nome eomes more varale, usng the standard, strtly onave utlty funton wth onstant relatve rsk averson, to stylze the fat that rsk delnes wth the level of nome and nreases wth varalty. * s+ n ρ ρ = y t n t= s y (3) where y* s rsk-adjusted nome, y s the nome of the perod, dentfes the household, and ρ s the oeffent of rsk averson. The latter was assgned a value of two for the alulaton. 5 Ths proedure downgrades the level of average nome otaned y an ndvdual or household through tme, when that average has resulted from a varale path. Wth regard to nome molty the seond of the stated ojetves ts ntensty n Argentna, and partularly ts varatons etween the phases dentfed, was analysed on the ass on household movements etween nome quntles. In other words, dstruton quntles were alulated for eah of the oservatons, and eah household s poston was dentfed n eah ase. It was then possle to dentfy dfferent paths. As noted aove, ths approah, whh analyses paths etween nome quantles an e alled nto queston, eause t fals to take aount of ntra-quantle movements, and also eause t treats paths nvolvng very dfferent hanges n nome n the same way. The approah was therefore omplemented y analysng orrelaton oeffents etween the household nomes otaned from the four oservatons, makng t possle to evaluate hanges n the postons and relatve dfferentals etween nome repents n the nome dstruton. The smaller the orrelaton, the larger the dfferenes etween the nomes otaned y the same households n two perods of tme, and, therefore, the greater the molty of nome. The Pearson and Spearman (rank) orrelatons were used for ths. To quantfy the nfluene of molty on the nome dstruton, the Shorroks adjustment of nequalty for molty ndex, mentoned aove, was alulated. 6 5 Estmatons made wth larger oeffents dd not alter the results otaned. 4

15 Fnally, we examned onvergent movements through the estmaton of regresson (7). V. The dataase used Inome nstalty and molty, along wth ther mpat on levels and hanges n the dstruton of nome, s s usually analysed employng longtudnal data,.e. data showng the dfferent nomes reeved y the same person or household through tme. Ths type of nformaton faes lmtatons, espeally due to the presene of attrton and reportng errors. The former feature.e. the progressve loss of unts of oservaton s due to varous reasons, suh as households leavng the panel or hangng address, or dffultes arsng n the feld work. It may e not-random and lead to a ased estmaton of molty. Reportng errors, even f also present n ross-seton data, are more mportant for omputng nome varalty; spefally, they may lead to over-estmaton of ther effet on nequalty. Gven these general lmtatons of longtudnal data and the partular lmtatons derved from the soure to e used here (see elow), esdes ths type of nformaton we also analyzed pseudo panel data n order to study households nome molty. The next sheme summarzes the advantages and lmtatons of eah of the two types of nformaton soures to e used n ths paper. Advantages Dsadvantages Longtudnal Data Control of heterogenety Attrton Greater measurement or reportng errors Pseudo Panels Control of measurement or reportng errors Extenson of the tme perod Heterogenety Generally they are not ndvdual data ut averages As the paper ams at explorng the mpat of laour market dynams, the unverse of households that was studed was lmted to those headed y ndvduals not over 65 years old. To otan results n terms of the nstalty of purhasng power, whh s the relevant onept, nomnal values were orreted for varatons n the onsumer pre ndex (CPI). 6 Felds and Sánhez Puerta (2005) and Alornoz and Menéndez (2002) address a smlar top to molty, y analysng the relaton etween the ntensty of the hange n nomes and ther level, usng the models represented n equaton (3). They also examne the relaton etween molty and nequalty. 5

16 5. Longtudnal data Although Argentna does not undertake longtudnal surveys, the permanent household survey (EPH), performed regularly y INDEC, 7 provdes data of that type that are useful for analysng these ssues. Longtudnal data an e omputed from the EPH eause ts sample panel s of a rotatng type: households are ntervewed on four suessve oasons. Consequently, y omparng the stuaton of an ndvdual n those four waves one an dedue the hanges experened n a numer varales, nludng nome and employment. EPH data were used showng the hanges experened n nomes, atvty status and oupaton for eah nome unt (ndvdual or household). Unts an also e haraterzed y a seres of soodemograph and employment attrutes. The EPH sample onssts of four rotaton groups, one of whh enters and another exts n eah of the two waves made eah year (n May and Otoer). On eah oason, therefore, 25% of the sample s renewed, so 75% of ases an e ompared etween two suessve waves. Aordngly, f one wanted to trak households for the maxmum possle tme,.e. durng the four waves n whh they reman n the survey durng an 8-month perod, t would only e possle to evaluate a suset representng 25% of the total sample. The proporton of households and ndvduals atually rentervewed s less than these amounts, however, eause ases fall y the waysde a degree of natural reduton (attrton) for varous reasons, suh as households leavng the panel or hangng address, or dffultes arsng n the feld work. As the sample sze was nsuffent, a ommonly used proedure was employed to onstrut ases y aggregatng (omnng) rotaton groups that entered the sample at dfferent ponts n tme. 8 Ths means that ndvduals (and households) who responded to the survey at dfferent tmes were onsdered smultaneously: n other words, the method aggregates hanges that ourred n neghourng ut dfferent perods. The data used refer to Greater Buenos Ares only, 9 sne mrodata are not avalale for the other zones nluded n the survey. In any event, the evoluton of the employment stuaton and nome dstruton n the metropoltan area has not dffered from that experened n other uran zones, so the onlusons to e reahed here may reasonaly e extrapolated to the whole set of regons. 20 To analyse nome paths, panel data were prepared for eah of the four stages dentfed. The followng sheme shows the dfferent rotaton groups for eah ase. Tale shows the numer of ndvduals who were employed at some pont n tme, and the households orrespondng to eah phase. 7 For a desrpton of the EPH methodology, see The survey sheme was hanged sustantally n Although ths proedure makes t possle to work wth a large numer of oservatons, the phenomenon of attrton an ntrodue sample ases whh have not yet een nvestgated. 9 Ths s Argentna s man uran agglomeraton, aountng for 30% of the ountry s populaton and 40% of ts total uran nhatants. 20 See, for example, Beara, Esquvel and Maurzo (2002). 6

17 Tale. Greater Buenos Ares: rotaton groups omprsng the sample n eah phase Phases Frst oservaton Seond oservaton Thrd oservaton Fourth oservaton Ot 987 May 988 Ot 988 May 989 May 988 Ot 988 May 989 Ot 989 Hgh nflaton Ot 988 May 989 Ot 989 May 990 May 989 Ot 989 May 990 Ot 990 Ot 989 May 990 Ot 990 May 99 May990 Ot 990 May 99 Ot 99 No. of ndvduals:,877 No. of households:,4 May 99 Ot 99 May 992 Ot 992 Ot 99 May 992 Ot 992 May 993 Stalzaton May 992 Ot 992 May 993 Ot 993 Ot 992 May 993 Ot 993 May 994 May 993 Ot 993 May 994 Ot 994 No. of ndvduals:,773 No. of households: 976 Ot 995 May 996 Ot 996 May 997 Reovery May 996 Ot 996 May 997 Ot 997 Ot 996 May 997 Ot 997 May 998 May 997 Ot 997 May 998 Ot 998 No. of ndvduals: 2,39 No. of households:,263 May 998 Ot 998 May 999 Ot 999 Ot 998 May 999 Ot 999 May-00 Reesson May 999 Ot 999 May 900 Ot 2000 Ot 999 May 2000 Ot 2000 May 200 May 2000 Ot 2000 May 200 Ot 200 No. of ndvduals: 3,29 No. of households:,65 Soure: Authors elaoraton ased on EPH. Comparng suessve waves of the survey underestmates the numer of hanges that atually ourred, eause transtons are eng dentfed y omparng two oservatons roughly sx months apart. Aordngly, ndvduals ould make two or more movements n the nterval etween the two waves (e.g. from natvty to unemployment and ve-versa), wthout these movements eng aptured. It should also e noted that the proedure only analyses the suset of nomes otaned y household memers as a result of ther laour-market partpaton as wage earners, ownaount workers or employers. Ths restrted defnton of nome faltates a learer relaton etween the dynams of nequalty and nome nstalty and the laour-market fators that appear as ther determnants. Household nome s measured y addng together the laour nomes of all employed memers. It also needs to e orne n mnd that the household survey used here, as s true of many others n the regon, does not adequately apture and sgnfantly under-reords the urrent resoures that households otan from ther ownershp of aptal. Evdene of ths s the smlarty of hanges reorded y ndators of nequalty n total and laour nome. 7

18 When studyng the nstalty of ndvdual laour nomes, the analyss nluded persons who were employed n at least one of the oservatons,.e. those who regstered some postve nome from employment. 5.2 Pseudo panels The use of pseudo panels allows one to follow the evoluton of ndvduals or households of a ertan ohort -usually defned y age- usng repeated ross-setonal data. In ths ase, we reated fttous ohorts of households y age of the household head, also usng data from the PHS. Hene, when the averages of two oservatons n the pseudo panel are ontrasted, what s atually eng ompared s the stuaton of groups of smlar haratersts ut not the same households. It has een shown that the nformaton omng from ths proedure has fewer lmtatons than longtudnal data and therefore t has een suggested that ths method would e more approprate for the analyss of nome varalty. 2 In order to otan an approprate numer of ases, the ohorts were grouped onsderng perods of fve years; thus, the households were grouped n eah ohort aordng to the fveyear perod n whh the age of the household head fell. We onsdered household surveys sne 984 due to greater avalalty of ross-setonal data. 22 For example, the ohort (see sheme elow) nludes those households that n eah Otoer-wave of the survey have household heads that were orn etween those years. The analyss was arred out groupng.e. makng a pool of- all the possle ohorts wth the avalale data of Greater Buenos Ares, whh are also shown n the followng tale. The study of nomes molty was arred out for the totalty of ohorts; however, for ertan analyses the household ohorts were dvded aordng to two eduatonal levels of the household head: nomplete hgh shool or less, and the rest. 23 As t wll e seen elow, ths proedure allows one to dstngush molty proesses etween sugroups of households that are homogeneous n ths varale whh onsttutes a reasonale varale to stratfy households. Numer of households n the sample Brth-year of the household s head Soure : Authors elaoraton ased on EPH. 2 See, for example, Antman and MKenze (2005) 22 The data ases from whh longtudnal nformaton an e otaned are avalale only from 987 onward. 23 It s worth to hghlght that smlar results were otaned when onsderng other eduatonal groupngs. 8

19 VI. Inome nstalty n Argentna n the 990s Ths seton addresses one of the artle s two ojetves, namely to study nome nstalty and ts effets on the level of welfare. The frst part analyses hanges n the degree of varalty of nomes, ther soures and the effet on dfferent groups of workers and households. The seond part revews the mpat on welfare and hanges theren durng the perod under analyss. 6.. Instalty of ndvdual and household nomes (a) Instalty of ndvdual nomes As shown n tale 2, there were no sgnfant hanges n the oeffent of varaton of laour nomes among ndvduals who were employed at some tme durng the four phases analysed. Ths result s urous eause, ontrary to expetatons, the sharp drop n nflaton that ourred etween the frst of those perods (overng the years efore the Convertlty At) and the other three dd not affet the average varalty of urrent nomes. As mentoned aove, nflaton nfluenes the varalty of an ndvdual s real laour nome through tme, va ts mpat on hanges n the remuneraton otaned n a gven jo. Thus, the drop n nflaton espeally from suh hgh rates as those reorded etween 987 and 99 helped to redue the nstalty of real wages. Ths s also shown n tale 2, whh evaluates nome varalty when the effets of jo nstalty are solated: the oeffent of varaton of laour remuneraton onsderng only postve nomes and exludng oservatons orrespondng to stuatons n whh the person was not employed falls n the seond perod ompared to the frst, and agan n the next one. 24 At the same tme, however, nome varalty assoated wth jo nstalty nreased, as an e dedued from the nrease n the oeffent of varaton of nomes, ontrollng for hanges n remuneraton. 24 As mentoned n note 6, the effets of varatons n remuneraton arsng from hanges of jo or hours worked y ndvduals that reman employed are not solated. Aordngly, nome varalty arsng from jo fators (ontrollng for varatons n real wages) ould e even greater f these effets are norporated. It should also e noted that an addtonal soure of nome varalty s data or measurement error n respet of delared nomes. An exerse that exluded ases n whh ths error may have ourred (.e. hanges n nome not assoated wth hanges of oupaton or hours worked) dd not alter the results otaned. 9

20 Tale 2 Greater Buenos Ares: oeffent of varaton of nomes of persons employed at least n one oservaton Indvduals under 65 years of age who were employed at some pont Lower Upper Total Atual Effet of varaton n real remuneratons (smulated ontrollng for jo nstalty) Effet of oupatonal varaton (smulated ontrollng for nstalty of remuneratons) Hgh nflaton phase Average Confdene nterval Stalzaton phase Confdene nterval Average Lower Upper Average Reovery phase Confdene nterval Lower Upper Reesson phase Confdene nterval Average Lower Upper Low-eduaton ndvduals Atual Effet of varaton n real remuneratons Effet of oupatonal varaton Medum-eduaton ndvduals Atual Effet of varaton n real remuneratons Effet of oupatonal varaton Hgh-eduaton ndvduals Atual Effet of varaton n real remuneratons Effet of oupatonal varaton Soure: Authors elaoraton ased on EPH. It an e onluded, therefore, that the average varalty of nomes among ndvduals who were employed at some pont remaned onstant throughout the perod, despte the sgnfant drop n nflaton aheved sne the early 990s. Ths does not mean that more stale pres have not had the expeted effets n terms of stalzng the purhasng power of remuneratons, ut those effets were ounterated y events n the laour market that nreased jo nstalty. Another way to verfy ths s to note that, n the late 980s, 69% of those who were employed at some pont had een employed n all four oservatons, whereas the proporton dereases y fve perentage ponts n the followng phase. Ths trend ontnued, alet less ntensvely, n the next two phases; and, as a result, the proporton of employed persons wth stale nome paths shrank y 0 perentage ponts etween the egnnng and end phases. A least-squares model was appled to evaluate the extent to whh ertan ndvdual and household varales were assoated wth nstalty of nome and ts oupatonal and remuneraton omponents. Status n terms of eduaton, 25 head of household, age and gender are onsdered usually to have a dret effet on nome varalty or some of ts determnants. 26 Tale 3 shows that all these attrutes were generally sgnfant and had the expeted sgns. Negatve and postve sgns are onfrmed n the ase of age and age squared, respetvely, refletng the expeted nfluene of the lfe yle: nstalty delnes as the age 25 Stratfaton aordng to eduatonal level was as follows: Low eduaton enompasses those wth nomplete seondary shoolng; medum eduaton represents those wth up to tertary eduaton nomplete; and hgh eduaton those that had ompleted the latter. In the least-squares analyss, a dummy varale was used for low eduaton, whh was gven the value of 0 for the medum and hgh eduaton levels. 26 There s ample evdene of the nfluene of those varales on jo nstalty; see, for example, Farer (999) and Nkel, Jones and Quntn (2000). 20

21 advanes, ut at a dereasng rate. Only when the dependent varale s pure remuneraton varalty, however, s the low eduaton oeffent not sgnfant n the frst perod, therey ndatng that the effets of nflaton were felt y the employed populaton at large. Ths stuaton was repeated followng the stalzaton of pres n the early 990s,.e. durng the stalzaton phase, whh also shows that the proess would have enefted all ndvduals, ndependently of other attrutes. In the other two phases, however, the low eduaton oeffent was sgnfant, suggestng that nstalty delned y less among suh ndvduals, or even nreased. Tale 3. Greater Buenos Ares: estmaton of fators determnng nome nstalty a Hgh nflaton phase Stalzaton phase Reovery phase Reesson phase Dependent varale: Effetve oeffent of varaton Low eduaton Male Age , Age squared Head Constant Dependent varale: Smulated oeffent of varaton, ontrollng for jo nstalty Low eduaton Male Age Age squared Head Constant Dependent varale: Smulated oeffent of varaton, ontrollng for remuneraton nstalty Low eduaton Male Age Age squared Head Constant Soure: Authors elaoraton ased on EPH. a Unless otherwse ndated, the oeffents are sgnfant at the 5% level. Not sgnfant at 5%. Not sgnfant at 0%. Among low-sklled employed persons, nome nstalty was greater towards the end of the perod analysed than at the start. Ths s shown dretly n tale 2 and an also e dedued from the model reported n tale 3 for eah of the four perods. 27 The sgnfane of ths result, however, emerges from an exerse (values not shown here), n whh a smlar model was appled to the set of oservatons n the four perods for low-eduaton ndvduals only, wth dummy varales representng the dfferent phases. The dummy varale 27 A smlar result was otaned y Gutérrez (2004) for the reessonary perod

22 orrespondng to the reessonary phase ( ) was postve and sgnfant wth respet to the frst phase (hgh nflaton, onsdered as the ase), ut ths was not the ase wth the those representng the other two phases. In ontrast, there were no sgnfant dfferenes when the exerse was repeated for hgher-eduaton ndvduals. Tale 2, aove, shows the nrease n nstalty for the low-eduaton group, whh remans unhanged for the other two groups. Ths proedure was also used to analyse the sgnfane of varatons n nome nstalty among the dfferent eduatonal groups, assoated ether wth flutuatons n remuneraton or wth oupatonal status. Among ndvduals wth lttle shoolng, oupatonal varalty was already nreasng at the start of the 990s whle pure nome nstalty was not hangng sgnfantly. In ontrast, the other group dd not show hanges n ether measure. As the foregong analyss shows, not only does the ndvdual nome varalty among people employed at some pont n tme dffer aordng to ther level of shoolng, ut ts persstene etween the phases analysed oneals dfferent ehavour patterns etween groups defned n ths way. The oeffents of varaton of nomes n the medum and hgh eduaton strata (tale 2) were roadly unhanged throughout the four perods, whereas the less eduated experened even greater nome nstalty n the fnal perod than n the hghnflaton phase. () Instalty of household nomes We now onsder the varalty of household nomes, whh s mportant not only for the analyss ut also to evaluate the extent to whh ths was affeted y the nstalty of laour nomes reeved y ndvduals who were employed at some pont. The relaton wll not neessarly e dret, sne t ould have een offset y the effet of other varales. Tale 4 shows a sgnfant derease n the oeffent of varaton (8%) of household laour nomes, etween the frst and seond perods n the early 990s, resultng from an nrease n average jo nstalty among households and a derease n the varalty of remuneraton. Then, durng the expansonary and reessonary phases that followed the mddle of the deade, famly nome varalty nreased agan (y 6% and 5% respetvely) on the ak of rsng jo nstalty. Nonetheless, the varalty of famly laour nomes n the last of the perods was 9% less than the value reorded n the late 980s; and although ths aggregate result onflates sgnfantly dfferent experenes aross strata defned y the eduaton level of household heads, on average t reflets a dfferent stuaton than for ndvdual nome varalty. 22

23 Tale 4. Greater Buenos Ares: oeffents of varaton of real laour nomes of households and numer of employed Households headed y persons under 65 years of age Lower Upper Employed memers Household laour nome Household laour nome ontrollng for jo nstalty Household laour nome ontrollng for remuneraton nstalty Households headed y persons under 65 years of age, wth low eduaton level Hgh nflaton phase Stalzaton phase Reovery phase Reesson phase Confdene Confdene Confdene Confdene nterval nterval nterval nterval Average Average Average Average Lower Upper Lower Upper Lower Upper Employed memers Household laour nome Household laour nome ontrollng for jo nstalty Household laour nome ontrollng for remuneraton nstalty Households headed y persons under 65 years of age, wth medum eduaton level Employed memers Household laour nome Household laour nome ontrollng for jo nstalty Household laour nome ontrollng for remuneraton nstalty Households headed y persons under 65 years of age, wth hgh eduaton level Employed memers Household laour nome Household laour nome ontrollng for jo nstalty Household laour nome ontrollng for remuneraton nstalty Soure: Authors elaoraton ased on EPH. The relevane of growng jo nstalty s revealed y a persstent rse n the oeffent of varaton of famly nomes alulated after ontrollng for hanges n varatons n the remuneraton of employed household memers (tale 4). Ths rose y 38% wth the stalzaton n the early 990s, whereas the pure real remuneraton hange dereased y 22%. Consderng the two end phases, however, the dfferenes etween the two measures were greater: nome varalty aused y jo nstalty nreased y 56%, whereas that stemmng from flutuatons n remuneratons was 8% elow the level reorded n the years of hgh nflaton. In keepng wth the analyss of ndvdual nomes, nome varalty assoated wth jo nstalty nreased most n households headed y ndvduals wth low levels of shoolng. It should e kept n mnd that the proedure used to measure varalty aused y jo hanges also aptures effets arsng from the strateges deployed y household memers n response to events affetng them. Spefally, susttuton and omplementarty mehansms operate among atve memers wthn households; and these affet nome nstalty through oth jos and remuneraton, wth the fnal outome dependng on whh effet prevals. 28 A lear example of ths s the hange n nome that an e assoated wth perfet susttuton of employed household memers (.e. f one memer eomes unemployed, another fnds a jo). If the nome of the new worker s dfferent than that of the famly memer who eomes unemployed, household nome s altered wthout any hange n the 28 See Beara and Grosman (2005). 23