Changes in the Structure of Family Income in Canada, The Effects of Changing Family Structure and Growing Proportion of Recent Immigrants

Transcription

1 NOT FOR CTATON WTHOUT AUTHOR PERMON Changes in he rucure of Family ncome in Canada The Effecs of Changing Family rucure and Growing Proporion of Recen mmigrans By Wen-Hao Chen Family and abour udies aisics Canada Oawa Onario K1A 0T6 Voice: Facsimile: nd Draf April 2003

2 Changes in he rucure of Family ncome in Canada : The Effecs of Changing Family rucure and Growing Proporion of Recen mmigrans Absrac The marke-based family income inequaliy in Canada has increased beween 1980 and 1997 paricularly in he 1990s. However unlike he case in he Unied aes where he increasing inequaliy was shown from a polarizaion a he wo ends of he income disribuion he growing inequaliy in Canada was largely associaed wih a major shif of densiy from he middle o he lower half of he income disribuion. Key possible explanaions of disappearing middle class are ofen associaed wih he increase in earnings dispersion including skill-biased echnological shocks open economy he growing correlaion of husband and wife earnings changes in family srucure or increasing low economic performance immigrans. Among possible explanaions affecing family income disribuion his paper focuses on he effecs of changing family srucure and he growing proporion of recen immigrans. The analyses are based on a condiional weighed densiy esimaion developed by DiNardo Forin and emieux 1996 and daa from urvey of Consumer Finance CF is used. The resuls show ha he increasing rend of non-radiional forms of family srucure had subsanial impac on rising family income inequaliy explaining one-fifh of he increase in Gini coefficien and one-hird of he growh in low-income rae. The impac of growing proporion of recen immigrans on income dispersion was also paramoun. is especially concenraed in he lower half of he disribuion explaining abou one-hird of he increase in and raios and was responsible for $462 of he decline in median income in his period.

3 . nroducion The increasing income polarizaion and povery have been he subjec of discussion in many counries. n he Unied aes he common wisdom of rising inequaliy has been associaed wih gains in boh he upper and lower porion of he income disribuion. n Canada marke-based family income inequaliy also grew rapidly over ime paricularly in he 1990s. 1 However he growing inequaliy in Canada was primarily associaed wih loss from he boom porion of he income disribuion. During he period beween 1980 and 1997 he real income index increased only 7 percen for people a he 90 h percenile of he income disribuion while i decreased abou 40 percen for hose a he 10 h percenile. n he meanime low-income rae calculaed as one half of he median increased abou 4.3 percenage poins his period despie he declining rend of median income. Wha has driven he change of family income srucure in Canada? Wha facors can explain he increase in inequaliy? The objecive of his paper is o undersand facors and heir conribuion o he changes in income disribuion and low-income rae. Among possible explanaions affecing family income disribuion his paper focuses on effecs of changing family srucure and he growing proporion of low economic performance of recen immigrans. 2 Previous Canadian lieraure has relaed rising inequaliy o various facors. For example McWaers and Beach 1990 provide an overview of conribuing facors affecing he disribuion of income including supply-side issues baby-boom effec 1. The definiion of family income hroughou his paper is oal marke income wihou governmen ransfers and ax componens from every member in he census family divided by an equivalence scale ha is defined as a square roo of family size. Noice ha inequaliy based on pos-ransfer family incomes has changed very lile over ime see also Beach and losve

4 increased female paricipaion changing family srucure reiremen and pensions and demand-side issues srucural changes and business cycle effecs. e lile is known abou he role of he growing proporion of recen immigrans on he increase in inequaliy. Recen immigrans are very differen from heir predecessors in erms of quaniy and qualiy. The grea volume of immigrans who came in he 1990s more likely came from Asia or developing counries where skills were less ransferable. hifs in skill disribuion e.g. language proficiency or credenial recogniion of recen immigrans aler heir earnings disribuion and possibly affec overall income inequaliy. The inclusion of he immigraion facor in he examinaion of changing income disribuion would help o fill he gap in he exising Canadian inequaliy lieraure. Besides invesigaing impacs of rising non-radiional families e.g. lone paren families coninue o be a crucial issue among policy agenda. The scope of governmen ransfers has been changed over ime wih a declining role in he 1990s. f he increase in single-head families was likely associaed wih low incomes more policy inervenions may be required. The analyses rely on a relaively new approach based on he condiional reweighing procedure developed by DiNardo Forin and emieux As noed by Daly and Vallea 2000 his echnique has wo advanages compared o he sandard regression-based decomposiion in he exising lieraure. Firs i enables esimaion of he enire densiy of income raher han jus mean or quiniles in he sandard regression model. econd i provides greaer behavioral conen on he resuls because he counerfacual disribuions consruced in his forma allow condiional relaionship beween explanaory facors and a se of oher aribues. 2. Recen immigrans are defined as hose who have been living in Canada 10 years or less. 2

5 The paper is organized as follows. ecion provides a brief lieraure review on inequaliy. ecion describes daa source and some hisorical rends in he daa. n secion V he condiional re-weighing decomposiion echnique applied in his paper will be described. The esimaed resuls are presened in secion V and a conclusion is given in secion V.. Facors Affecing Family ncome nequaliy: A ieraure Review Research on income inequaliy has been explored in grea deail hrough various perspecives. n he Unied aes discussions of rising polarizaion of family income disribuion have ofen focused on he widening male earnings dispersion driven by shifs in he relaive demand for skilled workers such as echnological changes Bound and Johnson 1992 flucuaions in he rae of growh of he supply of skilled workers Kaz and Murphy shifs in indusrial composiion ha are largely associaed wih rade or economic inegraion Murphy and Welch 1991 and changes in insiuional seings Podgursky 1983 Kaz and Murphy 1992 and DiNardo Forin and emieux n addiion o he earnings dispariies shifs in demographic composiion are also among he possible explanaions for he increase in family income inequaliy. For insance Karoly and Burless 1995 Burless 1999 and Daly and Vallea 2000 show ha he demographic changes especially for increasing in single head families are responsible for a grea proporion of he increase in overall inequaliy in he U. Oher sudies also indicae ha he increase in he female labour supply is largely associaed wih women whose husbands have relaively high earnings which would lead o an increase in family 3. Argumen on he echnological change and rising wage inequaliy is discussed in Card and DiNardo

6 income inequaliy Juhn and Murphy 1997 and Burless For example Burless 1999 found ha abou 13 percen of he rise in overall inequaliy beween 1979 and 1999 in he U can be aribued o an increase in he correlaion of husband and wife earned incomes. udies of income inequaliy are also well documened in Canada. On he issue of echnological change Gera Gu and in 2001 show ha a he aggregae level demand for skilled workers did no rise significanly over 1981 and 1994 he period ha echnological change was paramoun. Alhough hey also show ha he variaions are very subsanial across indusrial secors skill upgrading a he naional level is less eviden. Their resuls implicily sugges a less role of echnological change in rising income inequaliy. Wih respec o earnings dispariy Morissee 1995 shows ha he increase in earnings inequaliy in he 1980s no only occurred wih changes in earnings bu also in conjuncion wih changes in he disribuion of hours worked. Pico 1998 updaes he earnings inequaliy rends in he 1990s and finds ha he real wage raes have declined among young Canadians while he overall inequaliy has changed lile hrough 1980 and mid Wih respec o family income in Canada early sudy by Henderson and Rowley 1977 suggess ha he decline in family size is one of he major reasons for rising inequaliy since income variaion is much sronger wihin smaller families. Wolfson 1986 examines he imporance of shifing family srucure in income inequaliy. His resuls sugges ha he increasing incidence of lone paren families refleced by lower feriliy raes or rising separaion raes has enhanced he inequaliy over he period udies by Beach 1988 and McWaers and Beach 1990 discuss he changes in family income disribuion in relaion o he vanishing middle 4

7 class in he period beween he 1960s and 1980s. They examine many possible facors from boh supply side and demand side and conclude ha he declining share of middle class is largely associaed wih he decline in men s paricipaion rae and wih he rise in women s paricipaion rae while effecs from oher possible facors such as business cycle or indusrial shif are relaively small. Zyblock 1996 analyzes boh demographic and non-demographic facors on he change in family incomes over His resuls show ha in eiher age groups or family ypes he wihin group conribuion dominaes he beween group conribuion in he level of overall family income inequaliy. He also poins ou ha he shif oward lone paren families accouned for nearly 40 percen of he increase in inequaliy hrough his period. n addiion o he radiional houghs of growing earnings dispariies and increasing incidence of single head families his paper also address an imporan facor immigrans in relaion o he rising family income inequaliy in Canada. Recen sudy by Hou and Pico 2002 show ha aggregae low-income raes in Canada increased 2.3 percenage poins beween 1989 and 1999 and wo-hirds was accouned for by he rising low-income raes among immigrans. The effec was even sronger in regions wih high immigran populaion such as Torono and Vancouver.. Daa and Hisorical Trends This sudy used daa from he urvey of Consumer Finance CF for The sample includes individuals age 15 and above in he families bu excludes families whose major income recipiens are age 65 years and over because marke-based incomes are usually small among elderly families. The definiion of family in his sudy adops 5

8 he concep of census family which consiss of a married couple or common-law couple wihou or wih unmarried children or a lone paren wih an unmarried child or children. The major ineres in his sudy focuses on he changes of he disribuion of individual economic well-being based on marke force alone. To properly measure he well-being of an individual boh moneary and demographic componens in a family are considered. Family incomes are measured as annual marke incomes including earnings invesmen and privae ransfers bu excluding governmen ransfers and ax componens. The oal family marke incomes are furher sandardized hrough an equivalence scale in order o adjus demographic componens. n his conex equivalence scale is defined as he square roo of family size. The use of equivalence scale would reduce par of he inequaliy ha is caused by exacerbaion of big size families. n his seing individuals raher han families are considered as he basic uni of analysis. To define low-income saus individuals whose equivalen incomes fall below one half of he median equivalen income in a given year are considered as being in lowincome. For he decomposiion secion daa from wo peak years and one near peak year are used. The purpose of choosing hese hree paricular years aims o reduce any variaions ha were due o he business cycle effecs. All income measures are expressed in 2000 consan dollars. Figure 1 displays rends of inequaliy as indicaed by he percenile raio as well as low-income rae. No surprisingly he cyclical effecs are very obvious as shown by he wo spikes areas represening and recessions respecively. Focusing on he peak or near-peak years and 19 inequaliy 6

9 displays a growing up rend hroughou his period while low-income rae increased only during he 1990s. Boh indices increased subsanially in he early 1990s ye hey were never able o reurn o he previous level despie he economic expansions in he lae 1990s. Figure 2 shows ha he increasing inequaliy was mainly associaed wih he worse performance of people from he boom end of he income disribuion. The real income in 1997 for people a he 25 h percenile was abou 15 percen lower han he level in 1980 for people a he same percenile. The siuaion is even worse for hose exremely poor: he real income dropped as much as 40 percen for people a he 10 h percenile beween 1980 and On he conrary people a he op end of he disribuion experienced very lile change in erms of real income during his period. The real incomes for persons a he 75 h and 90 h percenile were only 4 and 7 percen respecively higher han hose a he same posiions in These paerns are very differen from hose found in he U lieraure where polarizaion is obvious e.g. Daly and Vallea n order o display changes in he enire income disribuion kernel densiy esimaes of he disribuion of equivalen income for peak years are shown in Figure 3. 4 Beween 1980 and 1997 he disribuion widened wih hollowing of he middle and densiy shif o boh ends. Densiy moved o he lef is more apparen suggesing he deerioraion of well-being for many individuals. The sory is very differen when wo periods were esimaed separaely. is surprising ha he change in income disribuion in he 1980s was mosly concenraed in he movemen from he middle porion o he 4. The kernel densiy is esimaed using he Gaussian funcional form and opimal bandwidh see ilverman 1986 for deail. 7

10 upper porion wih very lile change in he lower porion. On he conrary much of he densiy mass shifed o he lef in he 1990s while he change in he upper ail was relaively small. Figure 4 displays inequaliy measures for male earnings during his period. The measure includes people wih zero earnings. The rend of earnings dispersion was relaively sable in he 1980s and increased noably in he 1990s. ndeed he rising earnings dispersion possibly resuled from he decline in earnings among individuals in he lower porion of he disribuion as indicaed by a declining rend of median and less likely from gains in he upper porion of individuals. A he aggregae level i seems ha echnological changes or rade do no have much suppor o he changes in income disribuion in Canada. Neverheless heir effecs may be very significan wihin cerain groups or indusries Gera Gu and in Figure 5 depics he correlaion beween husband earnings and wife paricipaion. shows ha he increase in female paricipaion raes has been largely concenraed on women whose husbands have relaively high earnings. Among males whose earnings were righ on he op quarile of he disribuion he proporion of working wives increased significanly from.55 in 1980 o nearly.8 in On he oher hand he proporion of working wives remained relaively sable over ime among husbands whose earnings were on he boom quarile of he disribuion. The increasing correlaion beween husband earnings and wife labour force paricipaion is inclined o increase inequaliy in family income. Figure 6 displays he rend of family composiion. shows ha he Canadian family srucure has been changed wih a decline in husband-wife-children families 62% 8

11 in 1980 o 55% in 19 and a corresponding increase in all hree oher family ypes: combining 5.1 percenage poins increase in single unaached families and lone paren families and a sligh increase 2.1 percenage poins in childless couple families. The movemen oward single head families is believable o increase he proporion of lowincome families and herefore ends o increase inequaliy. The oher noable rend in Canada in he pas 20 years is he increasing proporion of recen immigrans and shifs of composiions among hem. According o CF recen immigrans accoun for abou 25 percen of he oal immigrans hroughou he 1980s bu heir share increased o 33 percen in Figure 7 shows ha new immigrans who came in he 1990s are more vulnerable han heir earlier cohors from he 1980s. n he 1983 recession he difference in median equivalen income beween recen immigrans and populaion was abou $4000 while he gap increased o $15000 in he 1993 recession. The decline in median income is likely refleced by a relaively weaker aachmen o full-ime employmen among new immigrans in he 1990s Figure 9. The proporion of full-ime employmen for recen immigrans sayed very close o he naional average abou.64 hroughou he 1980s hen plunged quickly in he 1990s and never came back. Of course he recession in he early 1990s played an imporan facor in his manner. Many sudies e.g. McDonald and Worswick 1998 have shown ha immigrans rae of earnings assimilaion is sensiive o macroeconomic condiions. However why are immigrans from he 1990s more negaively affeced by recession compared o heir earlier cohors from he 1980s? The sory behind he macroeconomic condiions is perhaps he change of composiion among immigrans. The skill disribuion of immigrans in he 1990s likely changed when he Third World counries 9

12 Asia Middle Eas and Africa replaced Europe and he Unied aes as he major source counries of immigrans. New immigrans in he 1990s are considered as less skilled han earlier cohors no necessarily because of he disadvanage in educaional aainmens bu because of problems in erms of credenial recogniion language proficiency and perhaps discriminaion. Figure 10 shows he proporion of immigrans who speak he official language as heir moher ongue. n 1981 over 45 percen of immigrans or recen immigrans spoke he official language as heir moher ongue. The percenage declined o 28 percen and 16 percen for immigrans and recen immigrans respecively in To wha exen he changes in he composiions of immigrans affec naional inequaliy or low-income rae is he cenral issue o be examined in his research. V. Mehodology The echnique for decomposing changes in he densiy of equivalen income is based upon he condiional re-weighing procedure developed by Dinardo Forin and emieux 1996 hereafer called DF. This echnique has also been used by many auhors in recen lieraure e.g. Vallea 2001 Daly and Vallea 2000 and Chiquiar and Hanson Basically i is similar o he famous Oaxaca-Blinder decomposiion. Bu his procedure is able o esimae he enire condiional disribuion while he Oaxaca-Blinder decomposiion only focused on he mean of a disribuion. The esimaed condiional weighs can be combined wih sample survey weighs o produce an adjused disribuion. Recall ha a sandard Oaxaca-Blinder decomposiion follows from an income equaion: 10

13 y y ˆ β * + * ˆ β ˆ β80 The lef-hand side is he change in average equivalen income y beween 1980 and equals o he sum of wo componens: ha due o differences in he mean characerisics and ha due o differences in he reurns o hese characerisics βˆ. The firs erm on he righ-hand side of 1 represens a counerfacual ha wha would he average equivalen income have been in year 1997 for individuals wih he mean characerisics of he 1980 levels. horcoming of his decomposiion is ha i only esimaes he mean of a disribuion and ignores changes for example a he ails. Disribuion changes oher han means such as quiniles or deciles are unable o be invesigaed wih his mehod. To consruc counerfacual densiies ha work wih he enire densiy of incomes DF shows ha i can be esimaed hrough kernel densiy esimaion o appropriaely weighed samples. f ˆ 1 y n n i 1 θ i y i K h h 2 Equaion 2 is a kernel densiy esimae based on a random sample... wih 1 n sampling weighs θ... θ h is he smoohing funcion referred as he bandwidh and 1 n K is he kernel funcion. 5 The choice of h and K may be sensiive o he disribuion and has been subjec o many discussions in he lieraure. n he conex he opimal bandwidh ilverman 1986 and Gaussian kernel funcion are used. To provide an exposiion of his echnique consider a simple binary variable moher ongue ha equals o 1 if immigrans speak he official language as heir moher ongue and zero oherwise. Wihou conrolling oher characerisics we have seen he 11

14 proporion of immigrans whose moher ongue is he official language has declined significanly over ime Figure 9. Therefore he simples way o impose he earlier disribuion of language profile on he curren equivalen income disribuion is o upweigh people who speak he official language as heir moher ongue by a number ha is equal o he percenage decrease in he share of his group over ime. imilarly a possibiliy o be in he group who speaks he non-official language as he moher ongue has increased. As a resul immigrans in his group will be downweigh because hey are less likely o be observed as such in earlier years han in curren. The condiional re-weighing funcion follows he spiri of he above simple example bu furher allows us o conrol oher characerisics. To begin wih consider is he disribuion of equivalen income in 1997 condiional on four explanaory facors immigrans full-ime employmen rae immigrans language and migraion duraion composiion family srucure and oher aribues o be examined in his conex. The disribuion of income can be shown as he produc of join densiies and : f f f 3 The challenge here is o consruc counerfacual densiies such ha wha would he densiy of equivalen income have been in 1997 if immigraion facors family srucure oher aribues bu nohing else had remained he same as in 1980 levels? f f ee ilverman 1986 for deailed on kernel densiy esimaion. 12

15 f 4 Wih proper arrangemen he counerfacual densiy in 4 can be expressed as f f 80 f f 5 which is equal o he exac densiy of equivalen income in 1997 imes four re-weighing funcions: and. The new weighs can hen be incorporaed ino he esimaion of he kernel densiy: n i i i h y K h n y f 1 1 ˆ θ 6 Esimaion of Re-weighing Funcions n his subsecion how he re-weighing funcions are esimaed in each case will be explained in more deail beginning wih he applicaion o immigraion facors. As seen in he previous secion labour marke performance among immigrans has been deerioraing since The declining full-ime employmen rae among immigrans is associaed wih many facors including observable composiional changes e.g. shif in moher ongue composiion and some unobservable changes e.g. credenials

16 discriminaion or macro condiions. To invesigae impacs of immigrans on he change of income disribuion he basic idea is o ask a quesion such as wha would he income disribuion be if labour marke performance in erms of full-ime employmen of immigrans condiional on oher observable characerisics e.g. age educaion family srucure geography moher ongue and duraion of residency ec. had remained a heir 1980 levels? This counerfacual densiy represens he overall effecs of immigrans on income disribuion due o he change in he full-ime employmen rae. could be due o any effecs ha are no conrolled here. For example shif of naional origin of immigrans o developing counries migh resul in more problems on credenials recogniion or discriminaion and herefore affec full-ime employmen probabiliy. Besides any oher changes such as increasing demands for cerain skills in favor of or no in favor of immigrans or changes in macro condiion are also capured in his caegory. n erms of noaion le be a binary variable wih value 1 if full-ime employmen and zero are oherwise. 80 Pr 1 80 Pr Pr 1 Pr 0 in 7 represens changes in he probabiliy beween 1980 and 1997 ha a given immigran defined by characerisics is observed o engage full-ime employmen. Every immigran in he disribuion will be eiher upweighed if 1 or downweighed if 1 in order o adjus full-ime employmen > < saus o is 1980 level. Esimae of he re-weighing funcion can be 14

17 obained by esimaing he condiional probabiliies in 7 hrough a probi model. For non-immigrans he re-weighing funcion is se o 1. n he firs decomposiion hold immigrans full-ime rae a he 1980 level condiional on heir moher ongue and migraion duraion saus. ndeed he proporion of recen immigrans has increased rapidly in he 1990s and fewer of hem speak he official language as heir moher ongue. is possible ha he economic reurns o hose who can speak he official language and o hose who possess longer Canadian experience were also changing in his period. As a resul wish o consruc a counerfacual densiy accouns for ineracion effecs ha holds consan he language/migraion duraion composiion in addiion o he full-ime employmen rae a heir 1980 levels. n a simple framework immigraion populaion can be divided ino four muually exclusive groups in a given year according o moher ongue 1 if moher ongue is official language 0 else and migraion duraion 1 if longer han 10 years 0 else. The re-weighing funcion for his facor is hen expressed as: 80 Pr Pr 1 + Pr Pr 3 4 c 1 c Pr c 2 Pr 2 Pr 2 4 Pr 4 Pr Pr c Esimae of he re-weighing in 8 can be obained hrough a mulinomial logi model. Again he re-weighing funcion is se o 1 for all non-immigrans. 15

18 imilarly esimae of he condiional re-weighing funcion for family srucure can also be calculaed by using mulinomial logi again because here are four differen family ypes as described in he ex. 80 Pr c c c 1 Pr c Finally applying Bayes rule he condiional re-weighing funcion for oher aribues can be wrien as: 80 Pr Pr 80 Pr 97 Pr is equal o he relaive probabiliy of observing people wih characerisics in he 1980 versus he 1997 sample imes he uncondiional probabiliies of being in eiher sample. The condiional probabiliies can be esimaed by probi model while he uncondiional probabiliies are simply he populaion raio. Changes in he densiy of equivalen income beween 1980 and 1997 can herefore be modeled based on he following decomposiion: f 97 f 80 f + f + f + f + f ; f ; 97 f ; ; 97 f 97 f ; 97 f 80 ; 97 ; 80 ; ; ; i ii iii iv v 16

19 The five componens in he above equaion represen he effecs of changing immigrans full-ime employmen rae immigrans language/migraion duraion composiions family srucure oher aribues and residual facors respecively. Criicism of his approach is ofen relaed o is inabiliy o disinguish overlapping effecs of differen facors. The possibiliy of a general equilibrium or an endogenous relaionship beween facors would confound he rue conribuion of each facor. A simple soluion is o perform he same decomposiion bu in differen order sequences and see wheher resuls are robus under a differen order. n his paper reverse-order decomposiion is also employed. V. Resuls This secion applies he decomposiion procedure described above o esimae he conribuion of wo immigraion facors family srucure oher aribues and residual on he change in equivalen income beween 1980 and Vecor of oher aribues includes age sex educaion province dummies size of area dummies and 16 Census Meropolian Area CMA dummies. Thus effecs such as aging populaion or shif in geographic locaion would be capured in his caegory. Finally facors such as skillbiased echnological shocks inernaional rade correlaion beween husband and wife earnings or oher unexplained facors fall ino he las caegory. Figures 10 o 13 display changes in he densiy of equivalen income beween 1980 and 1997 in decomposiion sequence. Each graph adjused an addiional modeled facor o is 1980 levels. The solid line in Figure 10 is he 1997 densiy of equivalen income while he dash line represens he counerfacual densiy ha adjused for he 17

20 immigrans full-ime employmen rae. shows ha he lower porion of he income disribuion would have been narrower slighly if full-ime employmen rae among immigrans are held consan a 1980 levels. The adjused disribuion moved densiy from he lef ail o he middle bu had virually no impac on he righ ail. Figure 11 furher adjused disribuion by holding immigrans language/migraion duraion composiions consan a he earlier year level. Changing hese wo composiions among immigrans had reduced densiy in he upper middle wih corresponding increase of mass in he lower middle of he disribuion. The visual effec became more significan when family srucure was added and held consan a he earlier year level Figure 12. Changing family srucure beween 1980 and 1997 had caused a uniform shif o he lef wih more concenraion in he lower ail of he disribuion. Finally he effec of changing oher aribues is displayed in Figure 13. Effecs such as shifs in age srucure facor or geographic locaion had resuled in a noable movemen in densiy from he lower half o he upper half of he disribuion. The quaniaive conribuions of visual presenaions are shown in Table 1. aisics include changes of a series of dispersion measures low-income rae and median income beween 1980 and Over his period inequaliy indices and lowincome rae based on equivalen income had increased. Overall inequaliy as measured by Gini coefficien increased abou 4.3 percenage poins. However he rising inequaliy is principally concenraed in he lower half of he income disribuion. Despie he declining median he change in raio is less noiceable bu he increase in raio is subsanial suggesing people in he lower end suffered major income loss over his period. 18

21 Turning o he conribuion of each facor wih respec o he changes in hese measures column 1 shows he effec of changing full-ime employmen rae of immigrans. Overall i explains 24 percen 26% of he increase in raio 12 percen of he increase in Gini coefficien and abou 17 percen or equivalenly 0.7 percenage poin of he increase in he low-income rae over his period. The sronger effecs on he and raios indicae ha increasing proporion of non-full ime jobs was likely associaed wih low incomes. Recall ha declining full-ime rae of immigrans was affeced by many unobserved facors. f shif of naional origin accouns for mos in his caegory i is equivalen o say ha he increasing immigrans from developing counries were likely associaed wih low incomes given all else being equal. Neverheless i is no idenifiable wih his daa. Column 2 furher displays he effec of immigrans due o he changes of composiions in moher ongue and migraion duraion. The proporion of recen immigrans increased significanly in he 1990s and fewer of hem speak he official language as heir moher ongue. By holding hese wo componens a he 1980 level i resuls in a 7 percen of increase in Gini coefficien 0.5 percenage poin growh in lowincome rae and $228 decline in median income. The magniudes are no considerably large in erms of overall changes. However i is surprising ha holding consan he proporion of recen immigrans and moher ongue composiion alone could resul in changes in inequaliy as well as low-income rae a he naional level. Combining 1 and 2 changes due o immigraion facors are subsanial. They accoun for 19 percen of he increase in Gini 33 percen of he increase in raio 1.2 percenage poins growh in low-income rae and $462 of he decline in median. 19

22 Changing family srucure is sill dominan among explanaory facors in explaining he change in equivalen income disribuion. Column 3 shows ha i accouns for 20 percen of he increase in Gini coefficien and raio and 32 percen or equivalenly 1.4 percenage poins of he growh in low-income rae. s impac was primarily concenraed in he lower half of he income disribuion; i explains 33 percen and 36 percen of he increase in and raios respecively while only 13 percen of he increase in raio. This suggess ha increasing single-head families lone parens or unaached families were associaed wih low-incomes. is also noable ha changes in family srucure were also responsible for as much as $755 decline in median over his period. As for he conribuion of oher aribues on he inequaliy measures column 4 shows he effecs are fairly small across all measures excep for movemen in median income. Acually facors in his caegory conribued o inequaliy in an opposie way. They brough down low-income rae by 0.7 percenage poin and also moved median up by $803 beween 1970 and is no sure which facor in his caegory conribued a subsanial offse effec in he growing inequaliy. One possible explanaion is he shif in age srucure. n 1997 he baby boom cohor had reached heir bes ime of life wih subsanial accumulaion of experience and human capial. is reasonable o see ha increasing proporion of prime-age people propelled he naional median upward. Finally he conribuion of residual facor is considerably large as lised in column 5 of Table 1. More han half of he increase in Gini and low-income rae remain unexplained. The residual effecs were even sronger in he upper half of he income disribuion compared o dispersions in he lower half. For insance more han 20

23 70 percen of he increase in he and raios canno be explained by any of hese four explanaory facors above while he shares are only 39 percen and 36 percen for he and raios respecively. The relaively large residual effecs in he upper ail dispersion preserve discussion of roles for oher unconrolled facors such as echnical globalizaion correlaion beween husband and wife earnings or perhaps simply he business cycle on he changes in equivalen income. Neverheless here is no way o disinguish such effecs from he residual in his conex. Reverse-Order Decomposiion As menioned earlier conribuions of explanaory facors may change if here were overlapping effecs beween facors. Therefore resuls from he reverse-order decomposiion are also presened in Table 2. Derivaion of re-weighing funcions for reverse order is described in Appendix A. Generally all measures display similar paerns as hose shown in he primary-order decomposiion. The conribuion of oher aribues wih respec o income dispersions excep for lower ail dispersion remained insignifican even if hey were considered as he firs in he decomposiion. neresingly he conribuion of raio has increased noiceably wih he expense from immigraion facors. is very likely overesimaed because facors such as a geographic shif from rural o urban areas is possibly associaed wih grea inflows of lowperformance immigrans who largely reside in ciies. Conribuions associaed wih changes in family srucure increased slighly in he reverse-order decomposiion noably in he upper half of he disribuion. For insance changes in family srucure accouned for 20 percen of he increase in he raio for 21

24 he reverse-order decomposiion while he share was only 13 percen for he primaryorder decomposiion in which he immigraion facors were condiioned. These effecs however migh be overesimaed because changing family srucure e.g. increasing single families was likely associaed wih he increase in young professional immigrans under he skill-based independen immigraion visa caegory. Finally effecs due o changes in immigrans facors reduced a bi when hey were considered he las in he decomposiion. Neverheless immigraion facors were sill responsible for 0.8 percenage poin of he increase in low-income rae and a noable $376 reducion in median income over his period. Decomposiion breakdown by wo periods: and Table 3 and Table 4 break down he previous analysis ino wo periods: and represening a peak-o-peak or near peak year in he 80s and 90s respecively. Peak-o-peak years are seleced for he purpose of avoiding possible variaions due o business cycles. urprisingly he increase in overall inequaliy was principally concenraed in he second period. More han wo-hirds of he increase in every inequaliy measure beween 1980 and 1997 ook place in he 1990s and here was no change in low-income rae in he firs period. n fac he enire economy was growing in he 1980s. The slow growh in inequaliy and a subsanial gain in median income $1441 sugges an overall improvemen of economy in he 1980s. The slighly increase of income dispersion in he lower half raio can be picked up mosly by he changes in family srucure. Neverheless i is worh noing ha he significan gain in median income in his period was mosly associaed wih residual facors migh be 22

25 echnical change or more open economy and par of he growh was acually offse by he immigraion facors as well as family srucure. The sory changed dramaically in he second period. Nearly 100 percen of he increase in low-income rae beween 1980 and 1997 occurred in he 1990s correspondingly wih a significan drop in median income $1900. Again residual or unexplained facors dominaed oher facors in explaining he change in inequaliy accouning for more han hree-quarers of he increase in mos of he measures in his period. Changes in family srucure were sill conribuing an influenial role paricularly in he lower half of he income dispersions explaining 25 percen of he increase in raio and 15 percen of he growh in he low-income rae during his period. One differen sory from he 1990s is ha he conribuion from immigraion facors became noiceable in he decomposiion. mmigraion facors from 1 and 2 of Table 4 ogeher explained 20 percen of he increase in Gini and nearly 1 percenage poin growh in low-income rae. The sudden increasing impac of immigrans on income disribuion is consisen wih he hisorical rend of increasing inflow of recen immigrans in he early 1990s. is also worh noing ha as much as a $1604 decline in median income and 74 percen or equivalenly 3.1 percenage poins of he increase in low-income rae remained unexplained during his period. ome migh relae hese o he declining role of governmen ransfers in he 1990s. is unlikely because governmen ransfers were excluded from he calculaion of equivalen income in his sudy. is also possible ha he overall decline in living sandards was jus reflecing he effec of a macro condiion. 23

26 Perhaps he economy in year 1997 was simply no good enough o compare o he economy in 1989 or V. Conclusion nequaliy based on equivalen income in Canada has increased beween 1980 and 1997 mainly in he 1990s. CF daa show ha inequaliy was largely associaed wih gains in he lower half raher han he increase in boh ends of he income disribuion. Wha has driven he changes of income srucure in Canada? Wha is he rising income dispersion in relaion o he increasing rend of non-radiional families? Do he increasing proporion of recen immigrans and changing composiions of immigrans affec inequaliy a he naional level? Wha policy implicaions have we learned from his? The objecive of his paper is o provide explanaions of facors ha affec inequaliy and low-income in Canada aking advanage of a new echnique developed by DiNardo Forin and emieux This procedure provides wo relaive advanages compared o he convenional regression-based or mean deviaion mehods in esimaing inequaliy. Firs i enables us o focus on he changes of he enire disribuion raher han mean or quiniles of a disribuion. econd he resuls provide more behaviour inerpreaion because counerfacual densiies consruced in his forma allow a condiional relaionship beween explanaory facors and a se of oher conrols. The analysis firs focused on he enire period from hen furher divided by wo sub-periods and which represen one peak-o-peak cycle and one peak-o-near peak business cycle. 24

27 During he enire period each of he four explanaory facors exhibied noable conribuions o he increase in inequaliy. The effecs are much sronger in explaining he increase in he lower half of he disribuion. They ogeher explained 64 percen and 61 percen of he and raio respecively. Bu he rising income dispersions in he upper half e.g raio were largely associaed wih unknown facors. Wih respec o he conribuion of each facor he effec of rising non-radiional families is consisen wih previous Canadian lieraure which would end o increase income inequaliy. Condiional on oher aribues changes in family srucure were responsible for 20 percen of he increase in Gini coefficien and 1.4 percenage poins increase in low-income rae beween 1980 and The impacs were especially salien on he lower half income dispersions suggesing ha increasing single-head families were more likely associaed wih low incomes. Changes in oher aribues had an equalizing effec on he disribuion of equivalen income. s counerfacual effec reduced low-income rae by 0.7 percenage poin and pushed he median $800 above. The growing people in he prime age due o he baby boom populaion as well as migraion from rural oward urban areas all worked o increase middle poin and migh explain he equalizing effec in his caegory. Turning o he immigraion facors surprisingly hey played quie an influenial role in explaining income dispersion. Holding consan full-ime employmen rae of immigrans a he earlier year level would reduce 26 percen of he increase in raio and drop low-income rae by 0.7 percenage poin. When proporion of recen immigrans and moher ongue composiions were held consan a he 1980 level i furher explained 0.5 percenage poin growh in low-income rae and 7 percen of he 25

28 increase in Gini coefficien. Condiional on family srucure and oher aribues immigraion facors explained abou 17 percen or equivalenly 1.2 percenage poins of he increase in low-income rae and were responsible for a $462 decline in median beween 1980 and Nearly all hese effecs ook place in he 1990s he period ha a grea inflow of a new ype of immigrans occurred. Finally i should be noed ha a grea proporion of he increase in inequaliy and low-income rae remained unexplained in boh periods. is no cerain which unknown facor dominaed he inequaliy rend in Canada in his sudy. n he U echnological change and open economy are common explanaions for he rising inequaliy or wage polarizaion. Raw daa from CF do no seem o suppor his view because he rising wage dispersion among males was no considerably large and he earnings gain from he upper porion of he disribuion was only modes. Alhough wage dispariy migh be very subsanial wihin some cerain groups or indusries hese effecs were washed ou in he aggregae level. The increasing correlaion beween husband and wife earnings was no examined in his conex. However raw daa do show ha here was a significan increase in he proporion of working wives among husbands whose earnings were on he op quarile of he disribuion. The increasing families wih wo high-wage earners migh be an ineresing issue o pursue in examining family income inequaliy in he fuure. Wha policy implicaions will we be focusing on from hese resuls? Decomposiion has shown ha marke-based inequaliy and low-income rae are highly associaed wih increasing single-head families and recen immigrans. raises he role of governmen ransfers as well as social programs. Whom should he welfare arge and 26

29 wha social programs e.g. career or language raining should be implemened as a prioriy? Do increasing credi consrains or minimum wages help single-head families and immigrans o improve heir marke oucome? How do we help skilled-immigrans o ge a connecion wih he hos counry? ssues such as discouning human capial of immigrans recogniion of foreign credenials discriminaion and increasing burden of he welfare sysem need o be aken ino accoun when discussing immigraion relaed policies in he fuure. gnoring such issues would lead o a more unequal sociey and perhaps social exclusion. 27

30 References Beach Charles M The Vanishing Middle Class?: Evidence and Explanaions. ndusrial Relaions Queen s Universiy Kingson. Beach Charles M and George A. losve Are We Becoming Two ocieies? ncome Polarizaion and he Myh of he Declining Meddle Class in Canada C.D. Howe nsiue Torono. Bound J. and G. Johnson Changes in he rucure of Wage in he 1980s: An Evaluaion of Alernaive Explanaions. American Economic Review 82 pp Burless Gary Effecs of Growing Wage Dispariies and Changing Family Composiion on he U.. ncome Disribuion. European Economic Review 43 pp Card D. and J. E. DiNardo kill-biased Technological Change and Rising wage nequaliy: ome Problems and Puzzles. Journal of abor Economics Vol. 20 No.4 pp Chiquiar Daniel and Gordon H. Hanson nernaional Migraion elf-elecion and he Disribuion of Wages: Evidence from Mexico and he Unied aes. Naional Bureau of Economic Research Working Paper 9242 Cambridge. Daly Mary C. and Rober G. Vallea nequaliy and Povery in he Unied aes: The Effecs of Changing Family Behavior and Rising Wage Dispersion. Research paper Economic Research Deparmen Federal Reserve Bank of an Francisco an Francisco. DiNardo John Nicole M. Forin and Thomas emieux abor Marke nsiuions and he Disribuion of Wages : A emiparameric Approach. Economerica 64: pp Gera urendra W. Gu and Z. in Technology and he Demand for kills in Canada: an ndusry-level Analysis. Canadian Journal of Economics Vol. 34 pp Henderson D. W. and J.C.R. Rowley The Disribuion and Evoluion of Canadian Family ncome Economic Council of Canada Discussion Paper #91. Juhn Chinhui and Kevin M. Murphy Wage nequaliy and Family abor upply. Journal of abor Economics 15 pp

31 Karoly ynn A. and Gary Burless Demographic Change Rising Earnings nequaliy and he Disribuion of Personal Well-Being Demography 323 pp Kaz. and K. Murphy Changes in Relaive Wages upply and Demand Facors. Quarerly Journal of Economics 107 pp McDonald James T. and C. Worswick The Earnings of mmigran men in Canada: Job enure cohor and macroeconomic condiions. ndusrial and abour Relaions Review 51 pp McWaers C. G. and C. M. Beach Facors Behind he Changes in Canada s Family ncome Disribuion and he hare of he Middle Class. ndusrial Relaions Vol. 45 Queen s Universiy Kingson. Morissee Rene Why Has nequaliy in Weekly Earnings ncreased in Canada? aisics Canada Analyical udies Branch Research Paper eries no 80. Murphy K. and F. Welch The Role of nernaional Trade in Wage Differenials. n Workers and Their Wages: Changing Paerns in he Unied aes ed. By M. Kosers. Washingon D.C.: American Enerprise nsiue. Oaxaca R Male-Female Wage Differenials in Urban abor Markes. nernaional Economic Review 14 pp Pico Garne Wha is Happening o Earnings nequaliy and ouh Wages in he 1990s? aisics Canada Analyical udies Branch Research Paper eries no 116. Pico Garne and Feng Hou The Rise in ow-ncome Among Recen mmigrans in Canada. aisics Canada Analyical udies Branch Research Paper eries no 198. Podgursky Michael Unions and Family ncome nequaliy. The Journal of Human Resources 38 v4 ppp ilverman B. W Densiy Esimaion for aisics and Daa Analysis. ondon: Chapman and Hall. Wolfson Michael C asis Amid Change: ncome nequaliy in Canada The Review of ncome and Wealh 32 no 4 pp Zyblock Myles Why is Family Marke ncome nequaliy ncreasing in Canada? Examining he Effecs of Aging Family Formaion Globalizaion and Technology. Applied Research Branch Working Paper Human Resources Developmen Canada. 29

32 Figure 1: Trends of nequaliy and ow-income Equivalen Marke ncome ow-income rae 75/25 percenile raio Real index year Figure 2: ndexed Equivalen ncome by percenile h percenile 75h percenile 10h percenile 25h percenile 1.4 Real income index year 30

33 Figure 3: Kernel Densiy for Equivalen ncome ear 1980 ear Equivalen ncome Figure 3A: Kernel Densiy for Equivalen ncome ear 1980 ear Equivalen ncome 31

34 Figure 3B: Kernel Densiy for Equivalen ncome ear 1989 ear Equivalen ncome Figure 4: nequaliy measures for Male earnings age Gini coefficien C.V. Median Real ndex year 32

35 Figure 5: Proporion of working wives by husband s quarile earnings disribuion couple families Boom quarile 3rd quarile 2nd quarile Top quarile.85 Proporion of working wives Figure 6: Family Composiion year ingle Couple wih children Couple wihou children one paren.65 Proporion of individuals age year.09 33

36 Figure 7: Median Equivalen Marke ncome age 15+ Populaion Recen mmigrans mmigrans Median equivalen income year Figure 8: Full-ime employmen rae ndividuals age All Recen immigrans mmigrans.75.7 Full-ime Employmen rae year 34

37 Figure 9: Proporion of immigrans whose moher ongue is English or French mmigrans Recen immigrans.5 Proporion of immigrans Figure 10: 1997 Equivalen ncomes wih 1980 mmigrans full-ime employmen year Family Equiv. ncome Holding mmigrans full-ime employmen rae Kernel Densiy Equivalen ncome 35

38 Figure 11: 1997 Equivalen ncomes wih 1980 mmigrans full-ime employmen moher ongue and migraion duraion composiions income wih 1980 immigrans full-ime employmen rae Adjused for immigrans moher ongue and migraion duraion composiions Kernel Densiy Equivalen ncome Figure 12: 1997 Equivalen ncomes wih 1980 mmigraion facors and family srucure income wih 1980 immigraion facors Adjused for family srucure Kernel Densiy Equivalen ncome 36